diff --git a/404.html b/404.html index 3a73d08..fa43ca8 100644 --- a/404.html +++ b/404.html @@ -50,7 +50,14 @@ - + diff --git a/LICENSE-text.html b/LICENSE-text.html index c72bd4f..b3c2de0 100644 --- a/LICENSE-text.html +++ b/LICENSE-text.html @@ -33,7 +33,13 @@ - + diff --git a/articles/index.html b/articles/index.html index 151920c..c3f01c8 100644 --- a/articles/index.html +++ b/articles/index.html @@ -33,7 +33,13 @@ - + diff --git a/articles/vignette.html b/articles/vignette.html index a299850..6655862 100644 --- a/articles/vignette.html +++ b/articles/vignette.html @@ -52,7 +52,14 @@ - + @@ -68,7 +75,7 @@

Anonymous

- + Source: vignettes/vignette.Rmd
vignette.Rmd
diff --git a/authors.html b/authors.html index e325ee7..69aabfc 100644 --- a/authors.html +++ b/authors.html @@ -33,7 +33,13 @@ - + @@ -54,7 +60,7 @@

Authors

Citation

-

+

Source: inst/CITATION

Thorson, J. T., Andrews, A. G., Essington, T., & Large, S. (In press). Dynamic structural equation models synthesize ecosystem dynamics constrained by ecological mechanisms. Methods in Ecology and Evolution.

@Article{,
diff --git a/index.html b/index.html
index 75b8555..f6d0bcc 100644
--- a/index.html
+++ b/index.html
@@ -60,7 +60,14 @@
         
 
 
-      
+
@@ -103,6 +110,8 @@

Dynamic structural equation modelsLinks

diff --git a/news/index.html b/news/index.html index 1025e76..7ef4339 100644 --- a/news/index.html +++ b/news/index.html @@ -33,7 +33,13 @@ - + @@ -41,11 +47,11 @@

Changelog

- + Source: NEWS.md
-

dsem 1.0.1

+

dsem 1.0.1

CRAN release: 2024-01-18

  • Fix bug in simulate.dsem to keep up with changing interface in dsem
  • Update CITATION to indicate accepted paper
    • diff --git a/pkgdown.yml b/pkgdown.yml index 4deb51f..2146d35 100644 --- a/pkgdown.yml +++ b/pkgdown.yml @@ -3,5 +3,5 @@ pkgdown: 2.0.7 pkgdown_sha: ~ articles: vignette: vignette.html -last_built: 2024-01-04T20:01Z +last_built: 2024-01-18T15:56Z diff --git a/reference/Rplot002.png b/reference/Rplot002.png index 34c2eb7..cb417f7 100644 Binary files a/reference/Rplot002.png and b/reference/Rplot002.png differ diff --git a/reference/TMBAIC.html b/reference/TMBAIC.html index c59ce0a..50bc5ee 100644 --- a/reference/TMBAIC.html +++ b/reference/TMBAIC.html @@ -33,7 +33,13 @@
-
    +
    @@ -41,7 +47,7 @@

    Calculate marginal AIC for a fitted model

    - + Source: R/TMBAIC.R
    TMBAIC.Rd
    diff --git a/reference/as_fitted_DAG.html b/reference/as_fitted_DAG.html index 43fe829..44f0926 100644 --- a/reference/as_fitted_DAG.html +++ b/reference/as_fitted_DAG.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Convert output from package dsem to phylopath

    - + Source: R/dsem.R
    as_fitted_DAG.Rd
    diff --git a/reference/as_sem.html b/reference/as_sem.html index fa3289a..9a5b659 100644 --- a/reference/as_sem.html +++ b/reference/as_sem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Convert dsem to sem output

    - + Source: R/dsem.R
    as_sem.Rd
    diff --git a/reference/bering_sea.html b/reference/bering_sea.html index dfaf527..c8bbb31 100644 --- a/reference/bering_sea.html +++ b/reference/bering_sea.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Bering Sea marine ecosystem

    - + Source: R/data.R
    bering_sea.Rd
    diff --git a/reference/classify_variables.html b/reference/classify_variables.html index f9aac46..332ac3e 100644 --- a/reference/classify_variables.html +++ b/reference/classify_variables.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Classify variables path

    - + Source: R/classify_variables.R
    classify_variables.Rd
    diff --git a/reference/dsem-1.png b/reference/dsem-1.png index b216880..8f1065e 100644 Binary files a/reference/dsem-1.png and b/reference/dsem-1.png differ diff --git a/reference/dsem-2.png b/reference/dsem-2.png index 7ad858b..e019f39 100644 Binary files a/reference/dsem-2.png and b/reference/dsem-2.png differ diff --git a/reference/dsem.html b/reference/dsem.html index 6059bd5..f047187 100644 --- a/reference/dsem.html +++ b/reference/dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Fit dynamic structural equation model

    - + Source: R/dsem.R
    dsem.Rd
    diff --git a/reference/dsem_control.html b/reference/dsem_control.html index 2d643fe..3e20838 100644 --- a/reference/dsem_control.html +++ b/reference/dsem_control.html @@ -37,7 +37,13 @@ -
    + @@ -45,7 +51,7 @@

    Detailed control for dsem structure

    - + Source: R/dsem.R
    dsem_control.Rd
    diff --git a/reference/fit_tmb.html b/reference/fit_tmb.html index 2247180..faf872b 100644 --- a/reference/fit_tmb.html +++ b/reference/fit_tmb.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Optimize a TMB model

    - + Source: R/fit_tmb.R
    fit_tmb.Rd
    diff --git a/reference/index.html b/reference/index.html index ce06ddf..4b24c50 100644 --- a/reference/index.html +++ b/reference/index.html @@ -33,7 +33,13 @@ -
    + diff --git a/reference/isle_royale.html b/reference/isle_royale.html index fd19239..0b1d65e 100644 --- a/reference/isle_royale.html +++ b/reference/isle_royale.html @@ -33,7 +33,13 @@ - + @@ -41,7 +47,7 @@

    Isle Royale wolf and moose

    - + Source: R/data.R
    isle_royale.Rd
    diff --git a/reference/list_parameters.html b/reference/list_parameters.html index 41df254..c097542 100644 --- a/reference/list_parameters.html +++ b/reference/list_parameters.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    List fixed and random effects

    - + Source: R/list_parameters.R
    list_parameters.Rd
    diff --git a/reference/logLik.dsem.html b/reference/logLik.dsem.html index ab78dfc..17fde7c 100644 --- a/reference/logLik.dsem.html +++ b/reference/logLik.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Marglinal log-likelihood

    - + Source: R/dsem.R
    logLik.dsem.Rd
    diff --git a/reference/make_dsem_ram.html b/reference/make_dsem_ram.html index 46e3a31..12d66a8 100644 --- a/reference/make_dsem_ram.html +++ b/reference/make_dsem_ram.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Make a RAM (Reticular Action Model)

    - + Source: R/make_dsem_ram.R
    make_dsem_ram.Rd
    diff --git a/reference/parse_path.html b/reference/parse_path.html index ae0cada..63415b5 100644 --- a/reference/parse_path.html +++ b/reference/parse_path.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Parse path

    - + Source: R/parse_path.R
    parse_path.Rd
    diff --git a/reference/plot.dsem.html b/reference/plot.dsem.html index 75341c0..5703b61 100644 --- a/reference/plot.dsem.html +++ b/reference/plot.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Simulate dsem

    - + Source: R/dsem.R
    plot.dsem.Rd
    diff --git a/reference/predict.dsem.html b/reference/predict.dsem.html index 949445c..c3411ca 100644 --- a/reference/predict.dsem.html +++ b/reference/predict.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    predictions using dsem

    - + Source: R/dsem.R
    predict.dsem.Rd
    diff --git a/reference/print.dsem.html b/reference/print.dsem.html index d140d74..e7e3b7f 100644 --- a/reference/print.dsem.html +++ b/reference/print.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Print fitted dsem object

    - + Source: R/dsem.R
    print.dsem.Rd
    diff --git a/reference/residuals.dsem.html b/reference/residuals.dsem.html index 256e677..2645371 100644 --- a/reference/residuals.dsem.html +++ b/reference/residuals.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Calculate residuals

    - + Source: R/dsem.R
    residuals.dsem.Rd
    diff --git a/reference/sea_otter.html b/reference/sea_otter.html index 00d5e9c..9db9ad5 100644 --- a/reference/sea_otter.html +++ b/reference/sea_otter.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Sea otter trophic cascade

    - + Source: R/data.R
    sea_otter.Rd
    diff --git a/reference/simulate.dsem.html b/reference/simulate.dsem.html index 2689f3d..b4f07bb 100644 --- a/reference/simulate.dsem.html +++ b/reference/simulate.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Simulate dsem

    - + Source: R/dsem.R
    simulate.dsem.Rd
    diff --git a/reference/summary.dsem.html b/reference/summary.dsem.html index 2bf1705..bfa8ddb 100644 --- a/reference/summary.dsem.html +++ b/reference/summary.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    summarize dsem

    - + Source: R/dsem.R
    summary.dsem.Rd
    diff --git a/reference/vcov.dsem.html b/reference/vcov.dsem.html index 7634490..b71bc6d 100644 --- a/reference/vcov.dsem.html +++ b/reference/vcov.dsem.html @@ -33,7 +33,13 @@ -
    + @@ -41,7 +47,7 @@

    Extract Variance-Covariance Matrix

    - + Source: R/dsem.R
    vcov.dsem.Rd
    diff --git a/search.json b/search.json index eb3afae..88d0dcb 100644 --- a/search.json +++ b/search.json @@ -1 +1 @@ -[{"path":"/articles/vignette.html","id":"comparison-with-dynamic-linear-models","dir":"Articles","previous_headings":"","what":"Comparison with dynamic linear models","title":"Demonstration of selected features","text":"first demonstrate dsem gives identical results dynlm well-known econometric model, Klein-1 model. Results show packages provide (almost) identical estimates standard errors. can also compare results using Laplace approximation obtained via numerical integration random effects using MCMC. example, MCMC results somewhat higher estimates exogenous variance parameters (presumably follow chi-squared distribution positive skewness), otherwise two produce similar estimates.","code":"data(KleinI, package=\"AER\") TS = ts(data.frame(KleinI, \"time\"=time(KleinI) - 1931)) # dynlm fm_cons <- dynlm(consumption ~ cprofits + L(cprofits) + I(pwage + gwage), data = TS) fm_inv <- dynlm(invest ~ cprofits + L(cprofits) + capital, data = TS) # fm_pwage <- dynlm(pwage ~ gnp + L(gnp) + time, data = TS) # dsem sem = \" # Link, lag, param_name cprofits -> consumption, 0, a1 cprofits -> consumption, 1, a2 pwage -> consumption, 0, a3 gwage -> consumption, 0, a3 cprofits -> invest, 0, b1 cprofits -> invest, 1, b2 capital -> invest, 0, b3 gnp -> pwage, 0, c2 gnp -> pwage, 1, c3 time -> pwage, 0, c1 \" tsdata = TS[,c(\"time\",\"gnp\",\"pwage\",\"cprofits\",'consumption', \"gwage\",\"invest\",\"capital\")] fit = dsem( sem=sem, tsdata = tsdata, newtonsteps = 0, estimate_delta0 = TRUE, control = dsem_control(quiet = TRUE) ) #> 1 regions found. #> Using 1 threads #> 1 regions found. #> Using 1 threads # Compile m1 = rbind( summary(fm_cons)$coef[-1,], summary(fm_inv)$coef[-1,], summary(fm_pwage)$coef[-1,] )[,1:2] m2 = summary(fit$opt$SD)[1:9,] m = rbind( data.frame(\"var\"=rownames(m1),m1,\"method\"=\"OLS\",\"eq\"=rep(c(\"C\",\"I\",\"Wp\"),each=3)), data.frame(\"var\"=rownames(m1),m2,\"method\"=\"GMRF\",\"eq\"=rep(c(\"C\",\"I\",\"Wp\"),each=3)) ) m = cbind(m, \"lower\"=m$Estimate-m$Std..Error, \"upper\"=m$Estimate+m$Std..Error ) # ggplot estimates longform = melt( as.data.frame(KleinI) ) longform$year = rep( time(KleinI), 9 ) p1 = ggplot( data=longform, aes(x=year, y=value) ) + facet_grid( rows=vars(variable), scales=\"free\" ) + geom_line( ) p2 = ggplot(data=m, aes(x=interaction(var,eq), y=Estimate, color=method)) + geom_point( position=position_dodge(0.9) ) + geom_errorbar( aes(ymax=as.numeric(upper),ymin=as.numeric(lower)), width=0.25, position=position_dodge(0.9)) # p3 = plot( as_fitted_DAG(fit) ) + expand_limits(x = c(-0.2,1) ) p4 = plot( as_fitted_DAG(fit, lag=1), text_size=4 ) + expand_limits(x = c(-0.2,1), y = c(-0.2,0) ) p1 p2 grid.arrange( arrangeGrob(p3, p4, nrow=2) ) library(tmbstan) # MCMC for both fixed and random effects mcmc = tmbstan( fit$obj, init=\"last.par.best\" ) summary_mcmc = summary(mcmc) # long-form data frame m1 = summary_mcmc$summary[1:17,c('mean','sd')] rownames(m1) = paste0( \"b\", seq_len(nrow(m1)) ) m2 = summary(fit$opt$SD)[1:17,c('Estimate','Std. Error')] m = rbind( data.frame('mean'=m1[,1], 'sd'=m1[,2], 'par'=rownames(m1), \"method\"=\"MCMC\"), data.frame('mean'=m2[,1], 'sd'=m2[,2], 'par'=rownames(m1), \"method\"=\"LA\") ) m$lower = m$mean - m$sd m$upper = m$mean + m$sd # plot ggplot(data=m, aes(x=par, y=mean, col=method)) + geom_point( position=position_dodge(0.9) ) + geom_errorbar( aes(ymax=as.numeric(upper),ymin=as.numeric(lower)), width=0.25, position=position_dodge(0.9)) #"},{"path":"/articles/vignette.html","id":"comparison-with-vector-autoregressive-models","dir":"Articles","previous_headings":"","what":"Comparison with vector autoregressive models","title":"Demonstration of selected features","text":"next demonstrate dsem gives similar results vector autoregressive (VAR) model. , analyze population abundance wolf moose populations Isle Royale 1959 2019, downloaded website (Vucetich, JA Peterson RO. 2012. population biology Isle Royale wolves moose: overview. URL: www.isleroyalewolf.org). dataset previously analyzed Chapter 14 User Manual R-package MARSS (Holmes, E. E., M. D. Scheuerell, E. J. Ward (2023) Analysis multivariate time-series using MARSS package. Version 3.11.8. NOAA Fisheries, Northwest Fisheries Science Center, 2725 Montlake Blvd E., Seattle, WA 98112, DOI: 10.5281/zenodo.5781847). , compare fits using dsem dynlm, well vector autoregressive model package vars, finally MARSS. Results show dsem can estimate parameters vector autoregressive model (VAM), exactly matches results vars, using dynlm, using MARSS.","code":"data(isle_royale) data = ts( log(isle_royale[,2:3]), start=1959) sem = \" # Link, lag, param_name wolves -> wolves, 1, arW moose -> wolves, 1, MtoW wolves -> moose, 1, WtoM moose -> moose, 1, arM \" # initial first without delta0 (to improve starting values) fit0 = dsem( sem = sem, tsdata = data, estimate_delta0 = FALSE, control = dsem_control(quiet=TRUE), getsd=FALSE ) #> Warning in nlminb(start = startpar, objective = fn, gradient = gr, control = #> nlminb.control, : NA/NaN function evaluation # Refit with delta0 fit = dsem( sem = sem, tsdata = data, estimate_delta0 = TRUE, control = dsem_control( quiet=TRUE, parameters = fit0$obj$env$parList()) ) # dynlm fm_wolf = dynlm( wolves ~ 1 + L(wolves) + L(moose), data=data ) # fm_moose = dynlm( moose ~ 1 + L(wolves) + L(moose), data=data ) # # MARSS library(MARSS) z.royale.dat <- t(scale(data.frame(data),center=TRUE,scale=FALSE)) royale.model.1 <- list( Z = \"identity\", B = \"unconstrained\", Q = \"diagonal and unequal\", R = \"zero\", U = \"zero\" ) kem.1 <- MARSS(z.royale.dat, model = royale.model.1) #> Success! algorithm run for 15 iterations. abstol and log-log tests passed. #> Alert: conv.test.slope.tol is 0.5. #> Test with smaller values (<0.1) to ensure convergence. #> #> MARSS fit is #> Estimation method: kem #> Convergence test: conv.test.slope.tol = 0.5, abstol = 0.001 #> Algorithm ran 15 (=minit) iterations and convergence was reached. #> Log-likelihood: -3.21765 #> AIC: 22.4353 AICc: 23.70964 #> #> Estimate #> B.(1,1) 0.8669 #> B.(2,1) -0.1240 #> B.(1,2) 0.1417 #> B.(2,2) 0.8176 #> Q.(X.wolves,X.wolves) 0.1341 #> Q.(X.moose,X.moose) 0.0284 #> x0.X.wolves 0.2324 #> x0.X.moose -0.6476 #> Initial states (x0) defined at t=0 #> #> Standard errors have not been calculated. #> Use MARSSparamCIs to compute CIs and bias estimates. SE <- MARSSparamCIs( kem.1 ) # Using VAR package library(vars) var = VAR( data, type=\"const\" ) ### Compile m1 = rbind( summary(fm_wolf)$coef[-1,], summary(fm_moose)$coef[-1,] )[,1:2] m2 = summary(fit$opt$SD)[1:4,] #m2 = cbind( \"Estimate\"=fit$opt$par, \"Std. Error\"=fit$opt$SD$par.fixed )[1:4,] m3 = cbind( SE$parMean[c(1,3,2,4)], SE$par.se$B[c(1,3,2,4)] ) colnames(m3) = colnames(m2) m4 = rbind( summary(var$varresult$wolves)$coef[-3,], summary(var$varresult$moose)$coef[-3,] )[,1:2] # Bundle m = rbind( data.frame(\"var\"=rownames(m1), m1, \"method\"=\"dynlm\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)), data.frame(\"var\"=rownames(m1), m2, \"method\"=\"dsem\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)), data.frame(\"var\"=rownames(m1), m3, \"method\"=\"MARSS\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)), data.frame(\"var\"=rownames(m1), m4, \"method\"=\"vars\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)) ) #knitr::kable( m1, digits=3) #knitr::kable( m2, digits=3) m = cbind(m, \"lower\"=m$Estimate-m$Std..Error, \"upper\"=m$Estimate+m$Std..Error ) # ggplot estimates ... interaction(x,y) causes an error sometimes library(ggplot2) library(ggpubr) library(ggraph) longform = reshape( isle_royale, idvar = \"year\", direction=\"long\", varying=list(2:3), v.names=\"abundance\", timevar=\"species\", times=c(\"wolves\",\"moose\") ) p1 = ggplot( data=longform, aes(x=year, y=abundance) ) + facet_grid( rows=vars(species), scales=\"free\" ) + geom_point( ) p2 = ggplot(data=m, aes(x=interaction(var,eq), y=Estimate, color=method)) + geom_point( position=position_dodge(0.9) ) + geom_errorbar( aes(ymax=as.numeric(upper),ymin=as.numeric(lower)), width=0.25, position=position_dodge(0.9)) # p3 = plot( as_fitted_DAG(fit, lag=1), rotation=0 ) + geom_edge_loop( aes( label=round(weight,2), direction=0)) + #arrow=arrow(), , angle_calc=\"along\", label_dodge=grid::unit(10,\"points\") ) expand_limits(x = c(-0.1,0) ) ggarrange( p1, p2, p3, labels = c(\"Time-series data\", \"Estimated effects\", \"Fitted path digram\"), ncol = 1, nrow = 3)"},{"path":"/articles/vignette.html","id":"multi-causal-ecosystem-synthesis","dir":"Articles","previous_headings":"","what":"Multi-causal ecosystem synthesis","title":"Demonstration of selected features","text":"next replicate analysis involving climate, forage fishes, stomach contents, recruitment predatory fish. results discussed paper describing dsem.","code":"data(bering_sea) Z = ts( bering_sea ) family = rep('fixed', ncol(bering_sea)) # Specify model sem = \" # Link, lag, param_name log_seaice -> log_CP, 0, seaice_to_CP log_CP -> log_Cfall, 0, CP_to_Cfall log_CP -> log_Esummer, 0, CP_to_E log_PercentEuph -> log_RperS, 0, Seuph_to_RperS log_PercentCop -> log_RperS, 0, Scop_to_RperS log_Esummer -> log_PercentEuph, 0, Esummer_to_Suph log_Cfall -> log_PercentCop, 0, Cfall_to_Scop log_SSB -> log_RperS, 0, SSB_to_RperS log_seaice -> log_seaice, 1, AR1, 0.001 log_CP -> log_CP, 1, AR2, 0.001 log_Cspring -> log_Cspring, 1, AR3, 0.001 log_Cfall -> log_Cfall, 1, AR4, 0.001 log_Esummer -> log_Esummer, 1, AR5, 0.001 log_SSB -> log_SSB, 1, AR6, 0.001 log_RperS -> log_RperS, 1, AR7, 0.001 log_PercentEuph -> log_PercentEuph, 1, AR8, 0.001 log_PercentCop -> log_PercentCop, 1, AR9, 0.001 \" # Fit fit = dsem( sem = sem, tsdata = Z, family = family, control = dsem_control(use_REML=FALSE, quiet=TRUE) ) #> Warning in nlminb(start = startpar, objective = fn, gradient = gr, control = #> nlminb.control, : NA/NaN function evaluation #> Warning in nlminb(start = startpar, objective = fn, gradient = gr, control = #> nlminb.control, : NA/NaN function evaluation ParHat = fit$obj$env$parList() # summary( fit ) # Timeseries plot oldpar <- par(no.readonly = TRUE) par( mfcol=c(3,3), mar=c(2,2,2,0), mgp=c(2,0.5,0), tck=-0.02 ) for(i in 1:ncol(bering_sea)){ tmp = bering_sea[,i,drop=FALSE] tmp = cbind( tmp, \"PSEM\"=ParHat$x_tj[,i] ) SD = as.list(fit$opt$SD,what=\"Std.\")$x_tj[,i] tmp = cbind( tmp, outer(tmp[,2],c(1,1)) + outer(ifelse(is.na(SD),0,SD),c(-1,1)) ) # plot( x=rownames(bering_sea), y=tmp[,1], ylim=range(tmp,na.rm=TRUE), type=\"p\", main=colnames(bering_sea)[i], pch=20, cex=2 ) lines( x=rownames(bering_sea), y=tmp[,2], type=\"l\", lwd=2, col=\"blue\", lty=\"solid\" ) polygon( x=c(rownames(bering_sea),rev(rownames(bering_sea))), y=c(tmp[,3],rev(tmp[,4])), col=rgb(0,0,1,0.2), border=NA ) } par(oldpar) # library(phylopath) library(ggplot2) library(ggpubr) library(reshape) library(gridExtra) longform = melt( bering_sea ) longform$year = rep( 1963:2023, ncol(bering_sea) ) p0 = ggplot( data=longform, aes(x=year, y=value) ) + facet_grid( rows=vars(variable), scales=\"free\" ) + geom_point( ) p1 = plot( (as_fitted_DAG(fit)), edge.width=1, type=\"width\", text_size=4, show.legend=FALSE, arrow = grid::arrow(type='closed', 18, grid::unit(10,'points')) ) + scale_x_continuous(expand = c(0.4, 0.1)) p1$layers[[1]]$mapping$edge_width = 1 p2 = plot( (as_fitted_DAG(fit, what=\"p_value\")), edge.width=1, type=\"width\", text_size=4, show.legend=FALSE, colors=c('black', 'black'), arrow = grid::arrow(type='closed', 18, grid::unit(10,'points')) ) + scale_x_continuous(expand = c(0.4, 0.1)) p2$layers[[1]]$mapping$edge_width = 0.5 #grid.arrange( arrangeGrob( p0+ggtitle(\"timeseries\"), # arrangeGrob( p1+ggtitle(\"Estimated path diagram\"), # p2+ggtitle(\"Estimated p-values\"), nrow=2), ncol=2 ) ) ggarrange(p1, p2, labels = c(\"Simultaneous effects\", \"Two-sided p-value\"), ncol = 1, nrow = 2)"},{"path":"/articles/vignette.html","id":"site-replicated-trophic-cascade","dir":"Articles","previous_headings":"","what":"Site-replicated trophic cascade","title":"Demonstration of selected features","text":"Finally, replicate analysis involving trophic cascade involving sea stars predators, sea urchin consumers, kelp producers. , results discussed paper describing dsem.","code":"data(sea_otter) Z = ts( sea_otter[,-1] ) # Specify model sem = \" Pycno_CANNERY_DC -> log_Urchins_CANNERY_DC, 0, x2 log_Urchins_CANNERY_DC -> log_Kelp_CANNERY_DC, 0, x3 log_Otter_Count_CANNERY_DC -> log_Kelp_CANNERY_DC, 0, x4 Pycno_CANNERY_UC -> log_Urchins_CANNERY_UC, 0, x2 log_Urchins_CANNERY_UC -> log_Kelp_CANNERY_UC, 0, x3 log_Otter_Count_CANNERY_UC -> log_Kelp_CANNERY_UC, 0, x4 Pycno_HOPKINS_DC -> log_Urchins_HOPKINS_DC, 0, x2 log_Urchins_HOPKINS_DC -> log_Kelp_HOPKINS_DC, 0, x3 log_Otter_Count_HOPKINS_DC -> log_Kelp_HOPKINS_DC, 0, x4 Pycno_HOPKINS_UC -> log_Urchins_HOPKINS_UC, 0, x2 log_Urchins_HOPKINS_UC -> log_Kelp_HOPKINS_UC, 0, x3 log_Otter_Count_HOPKINS_UC -> log_Kelp_HOPKINS_UC, 0, x4 Pycno_LOVERS_DC -> log_Urchins_LOVERS_DC, 0, x2 log_Urchins_LOVERS_DC -> log_Kelp_LOVERS_DC, 0, x3 log_Otter_Count_LOVERS_DC -> log_Kelp_LOVERS_DC, 0, x4 Pycno_LOVERS_UC -> log_Urchins_LOVERS_UC, 0, x2 log_Urchins_LOVERS_UC -> log_Kelp_LOVERS_UC, 0, x3 log_Otter_Count_LOVERS_UC -> log_Kelp_LOVERS_UC, 0, x4 Pycno_MACABEE_DC -> log_Urchins_MACABEE_DC, 0, x2 log_Urchins_MACABEE_DC -> log_Kelp_MACABEE_DC, 0, x3 log_Otter_Count_MACABEE_DC -> log_Kelp_MACABEE_DC, 0, x4 Pycno_MACABEE_UC -> log_Urchins_MACABEE_UC, 0, x2 log_Urchins_MACABEE_UC -> log_Kelp_MACABEE_UC, 0, x3 log_Otter_Count_MACABEE_UC -> log_Kelp_MACABEE_UC, 0, x4 Pycno_OTTER_PT_DC -> log_Urchins_OTTER_PT_DC, 0, x2 log_Urchins_OTTER_PT_DC -> log_Kelp_OTTER_PT_DC, 0, x3 log_Otter_Count_OTTER_PT_DC -> log_Kelp_OTTER_PT_DC, 0, x4 Pycno_OTTER_PT_UC -> log_Urchins_OTTER_PT_UC, 0, x2 log_Urchins_OTTER_PT_UC -> log_Kelp_OTTER_PT_UC, 0, x3 log_Otter_Count_OTTER_PT_UC -> log_Kelp_OTTER_PT_UC, 0, x4 Pycno_PINOS_CEN -> log_Urchins_PINOS_CEN, 0, x2 log_Urchins_PINOS_CEN -> log_Kelp_PINOS_CEN, 0, x3 log_Otter_Count_PINOS_CEN -> log_Kelp_PINOS_CEN, 0, x4 Pycno_SIREN_CEN -> log_Urchins_SIREN_CEN, 0, x2 log_Urchins_SIREN_CEN -> log_Kelp_SIREN_CEN, 0, x3 log_Otter_Count_SIREN_CEN -> log_Kelp_SIREN_CEN, 0, x4 # AR1 Pycno_CANNERY_DC -> Pycno_CANNERY_DC, 1, ar1 log_Urchins_CANNERY_DC -> log_Urchins_CANNERY_DC, 1, ar2 log_Otter_Count_CANNERY_DC -> log_Otter_Count_CANNERY_DC, 1, ar3 log_Kelp_CANNERY_DC -> log_Kelp_CANNERY_DC, 1, ar4 Pycno_CANNERY_UC -> Pycno_CANNERY_UC, 1, ar1 log_Urchins_CANNERY_UC -> log_Urchins_CANNERY_UC, 1, ar2 log_Otter_Count_CANNERY_UC -> log_Otter_Count_CANNERY_UC, 1, ar3 log_Kelp_CANNERY_UC -> log_Kelp_CANNERY_UC, 1, ar4 Pycno_HOPKINS_DC -> Pycno_HOPKINS_DC, 1, ar1 log_Urchins_HOPKINS_DC -> log_Urchins_HOPKINS_DC, 1, ar2 log_Otter_Count_HOPKINS_DC -> log_Otter_Count_HOPKINS_DC, 1, ar3 log_Kelp_HOPKINS_DC -> log_Kelp_HOPKINS_DC, 1, ar4 Pycno_HOPKINS_UC -> Pycno_HOPKINS_UC, 1, ar1 log_Urchins_HOPKINS_UC -> log_Urchins_HOPKINS_UC, 1, ar2 log_Otter_Count_HOPKINS_UC -> log_Otter_Count_HOPKINS_UC, 1, ar3 log_Kelp_HOPKINS_UC -> log_Kelp_HOPKINS_UC, 1, ar4 Pycno_LOVERS_DC -> Pycno_LOVERS_DC, 1, ar1 log_Urchins_LOVERS_DC -> log_Urchins_LOVERS_DC, 1, ar2 log_Otter_Count_LOVERS_DC -> log_Otter_Count_LOVERS_DC, 1, ar3 log_Kelp_LOVERS_DC -> log_Kelp_LOVERS_DC, 1, ar4 Pycno_LOVERS_UC -> Pycno_LOVERS_UC, 1, ar1 log_Urchins_LOVERS_UC -> log_Urchins_LOVERS_UC, 1, ar2 log_Otter_Count_LOVERS_UC -> log_Otter_Count_LOVERS_UC, 1, ar3 log_Kelp_LOVERS_UC -> log_Kelp_LOVERS_UC, 1, ar4 Pycno_MACABEE_DC -> Pycno_MACABEE_DC, 1, ar1 log_Urchins_MACABEE_DC -> log_Urchins_MACABEE_DC, 1, ar2 log_Otter_Count_MACABEE_DC -> log_Otter_Count_MACABEE_DC, 1, ar3 log_Kelp_MACABEE_DC -> log_Kelp_MACABEE_DC, 1, ar4 Pycno_MACABEE_UC -> Pycno_MACABEE_UC, 1, ar1 log_Urchins_MACABEE_UC -> log_Urchins_MACABEE_UC, 1, ar2 log_Otter_Count_MACABEE_UC -> log_Otter_Count_MACABEE_UC, 1, ar3 log_Kelp_MACABEE_UC -> log_Kelp_MACABEE_UC, 1, ar4 Pycno_OTTER_PT_DC -> Pycno_OTTER_PT_DC, 1, ar1 log_Urchins_OTTER_PT_DC -> log_Urchins_OTTER_PT_DC, 1, ar2 log_Otter_Count_OTTER_PT_DC -> log_Otter_Count_OTTER_PT_DC, 1, ar3 log_Kelp_OTTER_PT_DC -> log_Kelp_OTTER_PT_DC, 1, ar4 Pycno_OTTER_PT_UC -> Pycno_OTTER_PT_UC, 1, ar1 log_Urchins_OTTER_PT_UC -> log_Urchins_OTTER_PT_UC, 1, ar2 log_Otter_Count_OTTER_PT_UC -> log_Otter_Count_OTTER_PT_UC, 1, ar3 log_Kelp_OTTER_PT_UC -> log_Kelp_OTTER_PT_UC, 1, ar4 Pycno_PINOS_CEN -> Pycno_PINOS_CEN, 1, ar1 log_Urchins_PINOS_CEN -> log_Urchins_PINOS_CEN, 1, ar2 log_Otter_Count_PINOS_CEN -> log_Otter_Count_PINOS_CEN, 1, ar3 log_Kelp_PINOS_CEN -> log_Kelp_PINOS_CEN, 1, ar4 Pycno_SIREN_CEN -> Pycno_SIREN_CEN, 1, ar1 log_Urchins_SIREN_CEN -> log_Urchins_SIREN_CEN, 1, ar2 log_Otter_Count_SIREN_CEN -> log_Otter_Count_SIREN_CEN, 1, ar3 log_Kelp_SIREN_CEN -> log_Kelp_SIREN_CEN, 1, ar4 \" # Fit model fit = dsem( sem = sem, tsdata = Z, control = dsem_control(use_REML=FALSE, quiet=TRUE) ) # summary( fit ) # library(phylopath) library(ggplot2) library(ggpubr) get_part = function(x){ vars = c(\"log_Kelp\",\"log_Otter\",\"log_Urchin\",\"Pycno\") index = sapply( vars, FUN=\\(y) grep(y,rownames(x$coef))[1] ) x$coef = x$coef[index,index] dimnames(x$coef) = list( vars, vars ) return(x) } p1 = plot( get_part(as_fitted_DAG(fit)), type=\"width\", show.legend=FALSE) p1$layers[[1]]$mapping$edge_width = 0.5 p2 = plot( get_part(as_fitted_DAG(fit, what=\"p_value\" )), type=\"width\", show.legend=FALSE, colors=c('black', 'black')) p2$layers[[1]]$mapping$edge_width = 0.1 longform = melt( sea_otter[,-1], as.is=TRUE ) #> Warning in type.convert.default(X[[i]], ...): 'as.is' should be specified by #> the caller; using TRUE longform$X1 = 1999:2019[longform$X1] #> Warning in 1999:2019[longform$X1]: numerical expression has 1008 elements: only #> the first used longform$Site = gsub( \"log_Kelp_\", \"\", gsub( \"log_Urchins_\", \"\", gsub( \"Pycno_\", \"\", gsub( \"log_Otter_Count_\", \"\", longform$X2)))) longform$Species = sapply( seq_len(nrow(longform)), FUN=\\(i)gsub(longform$Site[i],\"\",longform$X2[i]) ) p3 = ggplot( data=longform, aes(x=X1, y=value, col=Species) ) + facet_grid( rows=vars(Site), scales=\"free\" ) + geom_line( ) ggarrange(p1 + scale_x_continuous(expand = c(0.3, 0)), p2 + scale_x_continuous(expand = c(0.3, 0)), labels = c(\"Simultaneous effects\", \"Two-sided p-value\"), ncol = 1, nrow = 2)"},{"path":"/authors.html","id":null,"dir":"","previous_headings":"","what":"Authors","title":"Authors and Citation","text":"James Thorson. Author, maintainer.","code":""},{"path":"/authors.html","id":"citation","dir":"","previous_headings":"","what":"Citation","title":"Authors and Citation","text":"Thorson, J. T., Andrews, . G., Essington, T., & Large, S. (press). Dynamic structural equation models synthesize ecosystem dynamics constrained ecological mechanisms. Methods Ecology Evolution.","code":"@Article{, title = {Dynamic structural equation models synthesize ecosystem dynamics constrained by ecological mechanisms}, journal = {Methods in Ecology and Evolution}, year = {2024}, author = {J. T. Thorson and A. G. Andrews and T. Essington and S. Large}, }"},{"path":"/index.html","id":"dynamic-structural-equation-models","dir":"","previous_headings":"","what":"Dynamic structural equation models","title":"Fit Dynamic Structural Equation Models","text":"Package dsem fits dynamic structural equation models, includes nested submodels: structural equation models vector autoregressive models dynamic factor analysis state-space autoregressive integrated moving average (ARIMA) models model several advantages: estimates direct, indirect, total effects among system variables, including simultaneous lagged effects recursive (cyclic) dependencies can estimate cumulative outcome press pulse experiments initial conditions differ stationary distribution system dynamics estimates structural linkages regression slopes jointly imputing missing values /measurement errors rapidly fitted Gaussian Markov random field (GMRF) Generalized Linear Mixed Model (GLMM), speed asymptotics associated allows granular control number parameters (restrictions parameters) used structure covariance among variables time, phylosem specifically intended minimal implementation, uses standard packages simplify input/output formatting: Input: time-series defined using class ts, NA missing values Input: structural trade-offs specified using syntax defined package sem Output: visualizing estimated trade-offs using igraph Output: access model output using standard S3-generic functions including summary, predict, residuals, simulate, AIC Please see package vignettes details regarding syntax features.","code":""},{"path":"/reference/TMBAIC.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate marginal AIC for a fitted model — TMBAIC","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"TMBAIC calculates AIC given model fit","code":""},{"path":"/reference/TMBAIC.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"","code":"TMBAIC(opt, k = 2, n = Inf)"},{"path":"/reference/TMBAIC.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"opt output nlminb optim k penalty additional fixed effects (default=2, AIC) n sample size, use AICc calculation (default=Inf, AICc=AIC)","code":""},{"path":"/reference/TMBAIC.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"AIC, parsimonious model AIC relative candidate models","code":""},{"path":"/reference/as_fitted_DAG.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert output from package dsem to phylopath — as_fitted_DAG","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"Convert dsem phylopath output","code":""},{"path":"/reference/as_fitted_DAG.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"","code":"as_fitted_DAG(fit, lag = 0, what = \"Estimate\", direction = 1)"},{"path":"/reference/as_fitted_DAG.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"fit Output dsem lag lag output whether output estimates =\"Estimate\", standard errors =\"Std_Error\" p-values =\"Std_Error\" direction whether include one-sided arrows direction=1, one- two-sided arrows direction=c(1,2)","code":""},{"path":"/reference/as_fitted_DAG.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"Convert output format supplied est_DAG","code":""},{"path":"/reference/as_sem.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert dsem to sem output — as_sem","title":"Convert dsem to sem output — as_sem","text":"Convert output package dsem sem","code":""},{"path":"/reference/as_sem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert dsem to sem output — as_sem","text":"","code":"as_sem(object, lag = 0)"},{"path":"/reference/as_sem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert dsem to sem output — as_sem","text":"object Output dsem lag lag extract visualize","code":""},{"path":"/reference/as_sem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Convert dsem to sem output — as_sem","text":"Convert output format supplied sem","code":""},{"path":"/reference/bering_sea.html","id":null,"dir":"Reference","previous_headings":"","what":"Bering Sea marine ecosystem — bering_sea","title":"Bering Sea marine ecosystem — bering_sea","text":"Data used demonstrate test ecosystem synthesis","code":""},{"path":"/reference/bering_sea.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Bering Sea marine ecosystem — bering_sea","text":"","code":"data(bering_sea)"},{"path":"/reference/classify_variables.html","id":null,"dir":"Reference","previous_headings":"","what":"Classify variables path — classify_variables","title":"Classify variables path — classify_variables","text":"classify_variables copied sem:::classifyVariables","code":""},{"path":"/reference/classify_variables.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Classify variables path — classify_variables","text":"","code":"classify_variables(model)"},{"path":"/reference/classify_variables.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Classify variables path — classify_variables","text":"model SEM model","code":""},{"path":"/reference/classify_variables.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Classify variables path — classify_variables","text":"Tagged-list defining exogenous endogenous variables","code":""},{"path":"/reference/classify_variables.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Classify variables path — classify_variables","text":"Copied package `sem` licence GPL (>= 2) permission John Fox","code":""},{"path":"/reference/dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit dynamic structural equation model — dsem","title":"Fit dynamic structural equation model — dsem","text":"Fits dynamic structural equation model","code":""},{"path":"/reference/dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit dynamic structural equation model — dsem","text":"","code":"dsem( sem, tsdata, family = rep(\"fixed\", ncol(tsdata)), estimate_delta0 = FALSE, control = dsem_control(), ... )"},{"path":"/reference/dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit dynamic structural equation model — dsem","text":"sem Specification time-series structural equation model structure including lagged simultaneous effects. See Details section make_dsem_ram description tsdata time-series data, outputted using ts family Character-vector listing distribution used column tsdata, element must fixed normal. family=\"fixed\" default behavior assumes given variable measured exactly. options correspond different specifications measurement error. estimate_delta0 Boolean indicating whether estimate deviations equilibrium initial year fixed effects, alternatively assume dynamics start stochastic draw away stationary distribution control Output dsem_control, used define user settings, see documentation function details. ... Additional parameters passed fit_tmb","code":""},{"path":"/reference/dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit dynamic structural equation model — dsem","text":"object (list) class `dsem`. Elements include: obj TMB object MakeADFun ram RAM parsed make_dsem_ram model SEM structure parsed make_dsem_ram intermediate description model linkages tmb_inputs list inputs passed MakeADFun opt output fit_tmb","code":""},{"path":"/reference/dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit dynamic structural equation model — dsem","text":"DSEM involves (minimum): Time series matrix \\(\\mathbf X\\) column \\(\\mathbf x_c\\) variable c time-series; Path diagram user-supplied specification path coefficients, define precision (inverse covariance) \\(\\mathbf Q\\) matrix state-variables see make_dsem_ram details math involved. model also estimates time-series mean \\( \\mathbf{\\mu}_c \\) variable. mean precision matrix therefore define Gaussian Markov random field \\(\\mathbf X\\): $$ \\mathrm{vec}(\\mathbf X) \\sim \\mathrm{MVN}( \\mathrm{vec}(\\mathbf{I_T} \\otimes \\mathbf{\\mu}), \\mathbf{Q}^{-1}) $$ Users can specify distribution measurement errors (assume variables measured without error) using argument family. defines link-function \\(g_c(.)\\) distribution \\(f_c(.)\\) time-series \\(c\\): $$ y_{t,c} \\sim f_c( g_c^{-1}( x_{t,c} ), \\theta_c )$$ dsem estimates specified coefficients, time-series means \\(\\mu_c\\), distribution measurement errors \\(\\theta_c\\) via maximizing log-marginal likelihood, also estimating state-variables \\(x_{t,c}\\). summary.dsem assembles estimates standard errors easy--read format. Standard errors fixed effects (path coefficients, exogenoux variance parameters, measurement error parameters) estimated matrix second derivatives log-marginal likelihod, standard errors random effects (.e., missing state-space variables) estimated generalization method (see sdreport details).","code":""},{"path":"/reference/dsem.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Fit dynamic structural equation model — dsem","text":"**Introducing package, features, comparison software (cite using dsem):** Thorson, J. T., Andrews, ., Essington, T., Large, S. (review). Dynamic structural equation models synthesize ecosystem dynamics constrained ecological mechanisms.","code":""},{"path":"/reference/dsem.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Fit dynamic structural equation model — dsem","text":"","code":"# Define model sem = \" # Link, lag, param_name cprofits -> consumption, 0, a1 cprofits -> consumption, 1, a2 pwage -> consumption, 0, a3 gwage -> consumption, 0, a3 cprofits -> invest, 0, b1 cprofits -> invest, 1, b2 capital -> invest, 0, b3 gnp -> pwage, 0, c2 gnp -> pwage, 1, c3 time -> pwage, 0, c1 \" # Load data data(KleinI, package=\"AER\") TS = ts(data.frame(KleinI, \"time\"=time(KleinI) - 1931)) tsdata = TS[,c(\"time\",\"gnp\",\"pwage\",\"cprofits\",'consumption', \"gwage\",\"invest\",\"capital\")] # Fit model fit = dsem( sem=sem, tsdata = tsdata, newtonsteps = 0, estimate_delta0 = TRUE, control = dsem_control(quiet=TRUE) ) #> NOTE: adding 8 variances to the model #> 1 regions found. #> Using 1 threads #> 1 regions found. #> Using 1 threads summary( fit ) #> path lag name start parameter first #> 1 cprofits -> consumption 0 a1 1 cprofits #> 2 cprofits -> consumption 1 a2 2 cprofits #> 3 pwage -> consumption 0 a3 3 pwage #> 4 gwage -> consumption 0 a3 3 gwage #> 5 cprofits -> invest 0 b1 4 cprofits #> 6 cprofits -> invest 1 b2 5 cprofits #> 7 capital -> invest 0 b3 6 capital #> 8 gnp -> pwage 0 c2 7 gnp #> 9 gnp -> pwage 1 c3 8 gnp #> 10 time -> pwage 0 c1 9 time #> 11 cprofits <-> cprofits 0 V[cprofits] 10 cprofits #> 12 consumption <-> consumption 0 V[consumption] 11 consumption #> 13 pwage <-> pwage 0 V[pwage] 12 pwage #> 14 gwage <-> gwage 0 V[gwage] 13 gwage #> 15 invest <-> invest 0 V[invest] 14 invest #> 16 capital <-> capital 0 V[capital] 15 capital #> 17 gnp <-> gnp 0 V[gnp] 16 gnp #> 18 time <-> time 0 V[time] 17 time #> second direction Estimate Std_Error z_value p_value #> 1 consumption 1 0.19323180 0.08199227 2.356708 1.843776e-02 #> 2 consumption 1 0.08942121 0.08136332 1.099036 2.717524e-01 #> 3 consumption 1 0.79625660 0.03594117 22.154444 9.451951e-109 #> 4 consumption 1 0.79625660 0.03594117 22.154444 9.451951e-109 #> 5 invest 1 0.48138159 0.08740017 5.507788 3.633710e-08 #> 6 invest 1 0.33084594 0.09069259 3.647993 2.642966e-04 #> 7 invest 1 -0.11150749 0.02408258 -4.630214 3.652878e-06 #> 8 pwage 1 0.44041578 0.02921826 15.073305 2.426629e-51 #> 9 pwage 1 0.14476508 0.03370695 4.294813 1.748404e-05 #> 10 pwage 1 0.13029843 0.02882713 4.519994 6.184145e-06 #> 11 cprofits 2 4.10929066 0.63141042 6.508114 7.610041e-11 #> 12 consumption 2 0.92285209 0.14239888 6.480754 9.126542e-11 #> 13 pwage 2 0.69302971 0.10701074 6.476263 9.402191e-11 #> 14 gwage 2 1.90952952 0.29464657 6.480746 9.127028e-11 #> 15 invest 2 0.91015014 0.14046555 6.479525 9.201155e-11 #> 16 capital 2 9.67951927 1.49358531 6.480727 9.128145e-11 #> 17 gnp 2 10.32018719 1.58101797 6.527558 6.685044e-11 #> 18 time 2 6.05530028 0.93435296 6.480742 9.127276e-11 plot( fit ) plot( fit, edge_label=\"value\" )"},{"path":"/reference/dsem_control.html","id":null,"dir":"Reference","previous_headings":"","what":"Detailed control for dsem structure — dsem_control","title":"Detailed control for dsem structure — dsem_control","text":"Define list control parameters. Note format input likely change rapidly dsem","code":""},{"path":"/reference/dsem_control.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Detailed control for dsem structure — dsem_control","text":"","code":"dsem_control( quiet = FALSE, run_model = TRUE, use_REML = TRUE, parameters = NULL, map = NULL )"},{"path":"/reference/dsem_control.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Detailed control for dsem structure — dsem_control","text":"quiet Boolean indicating whether run model printing messages terminal ; run_model Boolean indicating whether estimate parameters (default), instead return model inputs compiled TMB object without running; use_REML Boolean indicating whether treat non-variance fixed effects random, either motigate bias estimated variance parameters improve efficiency parameter estimation given correlated fixed random effects parameters list fixed random effects, e.g., constructed dsem modified hand (helpful advanced users change starting values restart intended values) map list fixed mirrored parameters, constructed dsem default available override default pass MakeADFun","code":""},{"path":"/reference/dsem_control.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Detailed control for dsem structure — dsem_control","text":"S3 object class \"dsem_control\" specifies detailed model settings, allowing user specification also specifying default values","code":""},{"path":"/reference/fit_tmb.html","id":null,"dir":"Reference","previous_headings":"","what":"Optimize a TMB model — fit_tmb","title":"Optimize a TMB model — fit_tmb","text":"fit_tmb runs TMB model generates standard diagnostics","code":""},{"path":"/reference/fit_tmb.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Optimize a TMB model — fit_tmb","text":"","code":"fit_tmb( obj, fn = obj$fn, gr = obj$gr, startpar = NULL, lower = -Inf, upper = Inf, getsd = TRUE, control = list(eval.max = 10000, iter.max = 10000, trace = 1), bias.correct = FALSE, bias.correct.control = list(sd = FALSE, split = NULL, nsplit = NULL, vars_to_correct = NULL), savedir = NULL, loopnum = 2, newtonsteps = 0, n = Inf, getReportCovariance = FALSE, getJointPrecision = FALSE, getHessian = FALSE, quiet = FALSE, start_time_elapsed = as.difftime(\"0:0:0\"), ... )"},{"path":"/reference/fit_tmb.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Optimize a TMB model — fit_tmb","text":"obj compiled TMB object fn Function minimized. Must return scalar value. . gr Optional function takes arguments objective evaluates gradient objective first argument. Must return vector long startpar startpar Starting values fixed effects (default NULL uses obj$par) lower, upper vectors lower upper bounds, replicated long start. unspecified, parameters assumed unconstrained. getsd Boolean indicating whether run standard error calculation; see sdreport details control list control parameters. details see nlminb bias.correct Boolean indicating whether epsilon bias-correction; see sdreport fit_tmbfor details bias.correct.control tagged list options epsilon bias-correction, vars_to_correct character-vector ADREPORT variables bias-corrected savedir directory save results (savedir=NULL, results saved) loopnum number times re-start optimization (loopnum=3 sometimes achieves lower final gradient loopnum=1) newtonsteps Integer specifying number extra newton steps take optimization (alternative loopnum). newtonstep requires calculating Hessian matrix therefore slow. well-behaved models, Newton step typically decrease maximum gradient loglikelihood respect fixed effect, therefore option can used achieve arbitrarily low final gradient given sufficient time well-behaved models. However, option also perform strangely unexpected consequences poorly-behaved models, e.g., fixed effects upper lower bounds. n sample sizes (n!=Inf n used calculate BIC AICc) getReportCovariance Get full covariance matrix ADREPORTed variables? getJointPrecision Optional. Return full joint precision matrix random effects parameters? getHessian return Hessian usage later code quiet Boolean whether print additional messages results terminal start_time_elapsed much time elapsed prior calling fit_tmb, use, e.g., calling fit_tmb multiple times sequence, start_time_elapsed = opt_previous$time_for_run ... list settings pass sdreport","code":""},{"path":"/reference/fit_tmb.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Optimize a TMB model — fit_tmb","text":"standard output nlminb, except additional diagnostics timing info, new slot containing output sdreport","code":""},{"path":"/reference/fit_tmb.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Optimize a TMB model — fit_tmb","text":"details see doi:10.1016/j.fishres.2015.11.016","code":""},{"path":"/reference/isle_royale.html","id":null,"dir":"Reference","previous_headings":"","what":"Isle Royale wolf and moose — isle_royale","title":"Isle Royale wolf and moose — isle_royale","text":"Data used demonstrate test cross-lagged (vector autoregressive) models","code":""},{"path":"/reference/isle_royale.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Isle Royale wolf and moose — isle_royale","text":"","code":"data(isle_royale)"},{"path":"/reference/isle_royale.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Isle Royale wolf and moose — isle_royale","text":"Data extracted file \"Data_wolves_moose_Isle_Royale_June2019.csv\" available https://isleroyalewolf.org/data/data/home.html obtained 2023-06-23. Reproduced permission John Vucetich, generated Wolves Moose Isle Royale project.","code":""},{"path":"/reference/isle_royale.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Isle Royale wolf and moose — isle_royale","text":"Vucetich, JA Peterson RO. 2012. population biology Isle Royale wolves moose: overview. https://www.isleroyalewolf.org","code":""},{"path":"/reference/list_parameters.html","id":null,"dir":"Reference","previous_headings":"","what":"List fixed and random effects — list_parameters","title":"List fixed and random effects — list_parameters","text":"list_parameters lists fixed random effects","code":""},{"path":"/reference/list_parameters.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"List fixed and random effects — list_parameters","text":"","code":"list_parameters(Obj, verbose = TRUE)"},{"path":"/reference/list_parameters.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"List fixed and random effects — list_parameters","text":"Obj Compiled TMB object verbose Boolean, whether print messages terminal","code":""},{"path":"/reference/list_parameters.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"List fixed and random effects — list_parameters","text":"Tagged-list fixed random effects, returned invisibly printed screen","code":""},{"path":"/reference/logLik.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Marglinal log-likelihood — logLik.dsem","title":"Marglinal log-likelihood — logLik.dsem","text":"Extract (marginal) log-likelihood dsem model","code":""},{"path":"/reference/logLik.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Marglinal log-likelihood — logLik.dsem","text":"","code":"# S3 method for dsem logLik(object, ...)"},{"path":"/reference/logLik.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Marglinal log-likelihood — logLik.dsem","text":"object Output dsem ... used","code":""},{"path":"/reference/logLik.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Marglinal log-likelihood — logLik.dsem","text":"object class logLik attributes val log-likelihood df number parameters Returns object class logLik. attributes \"df\" (degrees freedom) giving number (estimated) fixed effects model, abd \"val\" (value) giving marginal log-likelihood. class allows AIC work expected.","code":""},{"path":"/reference/make_dsem_ram.html","id":null,"dir":"Reference","previous_headings":"","what":"Make a RAM (Reticular Action Model) — make_dsem_ram","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"make_dsem_ram converts SEM arrow notation ram describing SEM parameters","code":""},{"path":"/reference/make_dsem_ram.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"","code":"make_dsem_ram(sem, times, variables, quiet = FALSE, remove_na = TRUE)"},{"path":"/reference/make_dsem_ram.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"sem Specification time-series structural equation model structure including lagged simultaneous effects. See Details section make_dsem_ram description times character vector listing set times order variables character vector listing set variables quiet Boolean indicating whether print messages terminal remove_na Boolean indicating whether remove NA values RAM (default) . remove_NA=FALSE might useful exploration diagnostics advanced users","code":""},{"path":"/reference/make_dsem_ram.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"reticular action module (RAM) describing dependencies","code":""},{"path":"/reference/make_dsem_ram.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"RAM specification using arrow--lag notation line RAM specification make_dsem_ram consists four (unquoted) entries, separated commas: 1. Arrow specification: simple formula, form -> B , equivalently, B <- regression coefficient (.e., single-headed directional arrow); <-> variance <-> B covariance (.e., double-headed bidirectional arrow). , B variable names model. name correspond observed variable, assumed latent variable. Spaces can appear freely arrow specification, can number hyphens arrows, including zero: Thus, e.g., ->B, --> B, >B legitimate equivalent. 2. Lag (using positive values): integer specifying whether linkage simultaneous (lag=0) lagged (e.g., X -> Y, 1, XtoY indicates X time T affects Y time T+1), one-headed arrows can lagged. Using positive values indicate lags matches notational convention used package dynlm. 3. Parameter name: name regression coefficient, variance, covariance specified arrow. Assigning name two arrows results equality constraint. Specifying parameter name NA produces fixed parameter. 4. Value: start value free parameter value fixed parameter. given NA (simply omitted), model provide default starting value. Lines may end comment following #. function extends code copied package `sem` licence GPL (>= 2) permission John Fox. Simultaneous autoregressive process simultaneous lagged effects text specifies linkages multivariate time-series model variables \\(\\mathbf X\\) dimensions \\(T \\times C\\) \\(T\\) times \\(C\\) variables. make_dsem_ram parses text build path matrix \\(\\mathbf{P}\\) dimensions \\(TC \\times TC\\), element \\(\\rho_{k_2,k_1}\\) represents impact \\(x_{t_1,c_1}\\) \\(x_{t_2,c_2}\\), \\(k_1=T c_1+t_1\\) \\(k_2=T c_2+t_2\\). path matrix defines simultaneous equation $$ \\mathrm{vec}(\\mathbf X) = \\mathbf P \\mathrm{vec}(\\mathbf X) + \\mathrm{vec}(\\mathbf \\Delta)$$ \\(\\mathbf \\Delta\\) matrix exogenous errors covariance \\(\\mathbf{V = \\Gamma \\Gamma}^t\\), \\(\\mathbf \\Gamma\\) Cholesky exogenous covariance. simultaneous autoregressive (SAR) process results \\(\\mathbf X\\) covariance: $$ \\mathrm{Cov}(\\mathbf X) = \\mathbf{(- P)}^{-1} \\mathbf{\\Gamma \\Gamma}^t \\mathbf{((- P)}^{-1})^t $$ Usefully, computing inverse-covariance (precision) matrix \\(\\mathbf{Q = V}^{-1}\\) require inverting \\(\\mathbf{(- P)}\\): $$ \\mathbf{Q} = (\\mathbf{\\Gamma}^{-1} \\mathbf{(- P)})^t \\mathbf{\\Gamma}^{-1} \\mathbf{(- P)} $$ Example: univariate first-order autoregressive model simultaneous autoregressive (SAR) process across variables times allows user specify simutanous effects (effects among variables within year \\(T\\)) lagged effects (effects among variables among years \\(T\\)). one example, consider univariate first-order autoregressive process \\(T=4\\). independent errors. specified passing sem = \"X -> X, 1, rho \\n X <-> X, 0, sigma\" make_dsem_ram. parsed RAM: Rows RAM heads=1 interpreted construct path matrix \\(\\mathbf P\\), column \"\" RAM indicates column number matrix, column \"\" RAM indicates row number matrix: $$ \\mathbf P = \\begin{bmatrix} 0 & 0 & 0 & 0 \\\\ \\rho & 0 & 0 & 0 \\\\ 0 & \\rho & 0 & 0 \\\\ 0 & 0 & \\rho & 0\\\\ \\end{bmatrix} $$ rows heads=2 interpreted construct Cholesky exogenous covariance \\(\\mathbf \\Gamma\\) column \"parameter\" RAM associates nonzero element two matrices element vector estimated parameters: $$ \\mathbf \\Gamma = \\begin{bmatrix} \\sigma & 0 & 0 & 0 \\\\ 0 & \\sigma & 0 & 0 \\\\ 0 & 0 & \\sigma & 0 \\\\ 0 & 0 & 0 & \\sigma\\\\ \\end{bmatrix} $$ two estimated parameters \\(\\mathbf \\beta = (\\rho, \\sigma) \\). results covariance: $$ \\mathrm{Cov}(\\mathbf X) = \\sigma^2 \\begin{bmatrix} 1 & \\rho^1 & \\rho^2 & \\rho^3 \\\\ \\rho^1 & 1 + \\rho^2 & \\rho^1 (1 + \\rho^2) & \\rho^2 (1 + \\rho^2) \\\\ \\rho^2 & \\rho^1 (1 + \\rho^2) & 1 + \\rho^2 + \\rho^4 & \\rho^1 (1 + \\rho^2 + \\rho^4) \\\\ \\rho^3 & \\rho^2 (1 + \\rho^2) & \\rho^1 (1 + \\rho^2 + \\rho^4) & 1 + \\rho^2 + \\rho^4 + \\rho^6 \\\\ \\end{bmatrix} $$ converges stationary covariance AR1 process times \\(t>>1\\): $$ \\mathrm{Cov}(\\mathbf X) = \\frac{\\sigma^2}{1+\\rho^2} \\begin{bmatrix} 1 & \\rho^1 & \\rho^2 & \\rho^3 \\\\ \\rho^1 & 1 & \\rho^1 & \\rho^2 \\\\ \\rho^2 & \\rho^1 & 1 & \\rho^1 \\\\ \\rho^3 & \\rho^2 & \\rho^1 & 1\\\\ \\end{bmatrix} $$ except lower pointwise variance initial times, arises \"boundary effect\". Similarly, arrow--lag notation can used specify SAR representing conventional structural equation model (SEM), cross-lagged (.k.. vector autoregressive) models (VAR), dynamic factor analysis (DFA), many time-series models.","code":""},{"path":"/reference/make_dsem_ram.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"","code":"# Univariate AR1 sem = \" X -> X, 1, rho X <-> X, 0, sigma \" make_dsem_ram( sem=sem, variables=\"X\", times=1:4 ) #> $model #> path lag name start parameter first second direction #> [1,] \"X -> X\" \"1\" \"rho\" NA \"1\" \"X\" \"X\" \"1\" #> [2,] \"X <-> X\" \"0\" \"sigma\" NA \"2\" \"X\" \"X\" \"2\" #> #> $ram #> heads to from parameter start #> 1 1 2 1 1 #> 2 1 3 2 1 #> 3 1 4 3 1 #> 5 2 1 1 2 #> 6 2 2 2 2 #> 7 2 3 3 2 #> 8 2 4 4 2 #> #> $variables #> [1] \"X\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # Univariate AR2 sem = \" X -> X, 1, rho1 X -> X, 2, rho2 X <-> X, 0, sigma \" make_dsem_ram( sem=sem, variables=\"X\", times=1:4 ) #> $model #> path lag name start parameter first second direction #> [1,] \"X -> X\" \"1\" \"rho1\" NA \"1\" \"X\" \"X\" \"1\" #> [2,] \"X -> X\" \"2\" \"rho2\" NA \"2\" \"X\" \"X\" \"1\" #> [3,] \"X <-> X\" \"0\" \"sigma\" NA \"3\" \"X\" \"X\" \"2\" #> #> $ram #> heads to from parameter start #> 1 1 2 1 1 #> 2 1 3 2 1 #> 3 1 4 3 1 #> 5 1 3 1 2 #> 6 1 4 2 2 #> 9 2 1 1 3 #> 10 2 2 2 3 #> 11 2 3 3 3 #> 12 2 4 4 3 #> #> $variables #> [1] \"X\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # Bivariate VAR sem = \" X -> X, 1, XtoX X -> Y, 1, XtoY Y -> X, 1, YtoX Y -> Y, 1, YtoY X <-> X, 0, sdX Y <-> Y, 0, sdY \" make_dsem_ram( sem=sem, variables=c(\"X\",\"Y\"), times=1:4 ) #> $model #> path lag name start parameter first second direction #> [1,] \"X -> X\" \"1\" \"XtoX\" NA \"1\" \"X\" \"X\" \"1\" #> [2,] \"X -> Y\" \"1\" \"XtoY\" NA \"2\" \"X\" \"Y\" \"1\" #> [3,] \"Y -> X\" \"1\" \"YtoX\" NA \"3\" \"Y\" \"X\" \"1\" #> [4,] \"Y -> Y\" \"1\" \"YtoY\" NA \"4\" \"Y\" \"Y\" \"1\" #> [5,] \"X <-> X\" \"0\" \"sdX\" NA \"5\" \"X\" \"X\" \"2\" #> [6,] \"Y <-> Y\" \"0\" \"sdY\" NA \"6\" \"Y\" \"Y\" \"2\" #> #> $ram #> heads to from parameter start #> 1 1 2 1 1 #> 2 1 3 2 1 #> 3 1 4 3 1 #> 5 1 6 1 2 #> 6 1 7 2 2 #> 7 1 8 3 2 #> 9 1 2 5 3 #> 10 1 3 6 3 #> 11 1 4 7 3 #> 13 1 6 5 4 #> 14 1 7 6 4 #> 15 1 8 7 4 #> 17 2 1 1 5 #> 18 2 2 2 5 #> 19 2 3 3 5 #> 20 2 4 4 5 #> 21 2 5 5 6 #> 22 2 6 6 6 #> 23 2 7 7 6 #> 24 2 8 8 6 #> #> $variables #> [1] \"X\" \"Y\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # Dynamic factor analysis with one factor and two manifest variables # (specifies a random-walk for the factor, and miniscule residual SD) sem = \" factor -> X, 0, loadings1 factor -> Y, 0, loadings2 factor -> factor, 1, NA, 1 X <-> X, 0, NA, 0.01 # Fix at negligible value Y <-> Y, 0, NA, 0.01 # Fix at negligible value \" make_dsem_ram( sem=sem, variables=c(\"X\",\"Y\",\"factor\"), times=1:4 ) #> NOTE: adding 1 variances to the model #> $model #> path lag name start parameter first second #> [1,] \"factor -> X\" \"0\" \"loadings1\" NA \"1\" \"factor\" \"X\" #> [2,] \"factor -> Y\" \"0\" \"loadings2\" NA \"2\" \"factor\" \"Y\" #> [3,] \"factor -> factor\" \"1\" NA \"1\" \"0\" \"factor\" \"factor\" #> [4,] \"X <-> X\" \"0\" NA \"0.01\" \"0\" \"X\" \"X\" #> [5,] \"Y <-> Y\" \"0\" NA \"0.01\" \"0\" \"Y\" \"Y\" #> [6,] \"factor <-> factor\" \"0\" \"V[factor]\" NA \"3\" \"factor\" \"factor\" #> direction #> [1,] \"1\" #> [2,] \"1\" #> [3,] \"1\" #> [4,] \"2\" #> [5,] \"2\" #> [6,] \"2\" #> #> $ram #> heads to from parameter start #> 1 1 1 9 1 #> 2 1 2 10 1 #> 3 1 3 11 1 #> 4 1 4 12 1 #> 5 1 5 9 2 #> 6 1 6 10 2 #> 7 1 7 11 2 #> 8 1 8 12 2 #> 9 1 10 9 0 1 #> 10 1 11 10 0 1 #> 11 1 12 11 0 1 #> 13 2 1 1 0 0.01 #> 14 2 2 2 0 0.01 #> 15 2 3 3 0 0.01 #> 16 2 4 4 0 0.01 #> 17 2 5 5 0 0.01 #> 18 2 6 6 0 0.01 #> 19 2 7 7 0 0.01 #> 20 2 8 8 0 0.01 #> 21 2 9 9 3 #> 22 2 10 10 3 #> 23 2 11 11 3 #> 24 2 12 12 3 #> #> $variables #> [1] \"X\" \"Y\" \"factor\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # ARIMA(1,1,0) sem = \" factor -> factor, 1, rho1 # AR1 component X -> X, 1, NA, 1 # Integrated component factor -> X, 0, NA, 1 X <-> X, 0, NA, 0.01 # Fix at negligible value \" make_dsem_ram( sem=sem, variables=c(\"X\",\"factor\"), times=1:4 ) #> NOTE: adding 1 variances to the model #> $model #> path lag name start parameter first second #> [1,] \"factor -> factor\" \"1\" \"rho1\" NA \"1\" \"factor\" \"factor\" #> [2,] \"X -> X\" \"1\" NA \"1\" \"0\" \"X\" \"X\" #> [3,] \"factor -> X\" \"0\" NA \"1\" \"0\" \"factor\" \"X\" #> [4,] \"X <-> X\" \"0\" NA \"0.01\" \"0\" \"X\" \"X\" #> [5,] \"factor <-> factor\" \"0\" \"V[factor]\" NA \"2\" \"factor\" \"factor\" #> direction #> [1,] \"1\" #> [2,] \"1\" #> [3,] \"1\" #> [4,] \"2\" #> [5,] \"2\" #> #> $ram #> heads to from parameter start #> 1 1 6 5 1 #> 2 1 7 6 1 #> 3 1 8 7 1 #> 5 1 2 1 0 1 #> 6 1 3 2 0 1 #> 7 1 4 3 0 1 #> 9 1 1 5 0 1 #> 10 1 2 6 0 1 #> 11 1 3 7 0 1 #> 12 1 4 8 0 1 #> 13 2 1 1 0 0.01 #> 14 2 2 2 0 0.01 #> 15 2 3 3 0 0.01 #> 16 2 4 4 0 0.01 #> 17 2 5 5 2 #> 18 2 6 6 2 #> 19 2 7 7 2 #> 20 2 8 8 2 #> #> $variables #> [1] \"X\" \"factor\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # ARIMA(0,0,1) sem = \" factor -> X, 0, NA, 1 factor -> X, 1, rho1 # MA1 component X <-> X, 0, NA, 0.01 # Fix at negligible value \" make_dsem_ram( sem=sem, variables=c(\"X\",\"factor\"), times=1:4 ) #> NOTE: adding 1 variances to the model #> $model #> path lag name start parameter first second #> [1,] \"factor -> X\" \"0\" NA \"1\" \"0\" \"factor\" \"X\" #> [2,] \"factor -> X\" \"1\" \"rho1\" NA \"1\" \"factor\" \"X\" #> [3,] \"X <-> X\" \"0\" NA \"0.01\" \"0\" \"X\" \"X\" #> [4,] \"factor <-> factor\" \"0\" \"V[factor]\" NA \"2\" \"factor\" \"factor\" #> direction #> [1,] \"1\" #> [2,] \"1\" #> [3,] \"2\" #> [4,] \"2\" #> #> $ram #> heads to from parameter start #> 1 1 1 5 0 1 #> 2 1 2 6 0 1 #> 3 1 3 7 0 1 #> 4 1 4 8 0 1 #> 5 1 2 5 1 #> 6 1 3 6 1 #> 7 1 4 7 1 #> 9 2 1 1 0 0.01 #> 10 2 2 2 0 0.01 #> 11 2 3 3 0 0.01 #> 12 2 4 4 0 0.01 #> 13 2 5 5 2 #> 14 2 6 6 2 #> 15 2 7 7 2 #> 16 2 8 8 2 #> #> $variables #> [1] \"X\" \"factor\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\""},{"path":"/reference/parse_path.html","id":null,"dir":"Reference","previous_headings":"","what":"Parse path — parse_path","title":"Parse path — parse_path","text":"parse_path copied sem::parse.path","code":""},{"path":"/reference/parse_path.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Parse path — parse_path","text":"","code":"parse_path(path)"},{"path":"/reference/parse_path.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Parse path — parse_path","text":"path text parse","code":""},{"path":"/reference/parse_path.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Parse path — parse_path","text":"Tagged-list defining variables direction specified path coefficient","code":""},{"path":"/reference/parse_path.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Parse path — parse_path","text":"Copied package `sem` licence GPL (>= 2) permission John Fox","code":""},{"path":"/reference/plot.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate dsem — plot.dsem","title":"Simulate dsem — plot.dsem","text":"Plot fitted dsem model","code":""},{"path":"/reference/plot.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate dsem — plot.dsem","text":"","code":"# S3 method for dsem plot(x, y, edge_label = c(\"name\", \"value\"), digits = 2, ...)"},{"path":"/reference/plot.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate dsem — plot.dsem","text":"x Output dsem y used edge_label Whether plot parameter names estimated values digits integer indicating number decimal places used ... arguments passed plot.igraph","code":""},{"path":"/reference/plot.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate dsem — plot.dsem","text":"Invisibly returns output graph_from_data_frame passed plot.igraph plotting.","code":""},{"path":"/reference/plot.dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate dsem — plot.dsem","text":"function coerces output graph plots graph.","code":""},{"path":"/reference/predict.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"predictions using dsem — predict.dsem","title":"predictions using dsem — predict.dsem","text":"Predict variables given new (counterfactual) values data, future past times","code":""},{"path":"/reference/predict.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"predictions using dsem — predict.dsem","text":"","code":"# S3 method for dsem predict(object, newdata = NULL, type = c(\"link\", \"response\"), ...)"},{"path":"/reference/predict.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"predictions using dsem — predict.dsem","text":"object Output dsem newdata optionally, data frame look variables predict. omitted, fitted data used create predictions. desiring predictions fitted data, user must append rows NAs future times. Similarly, desiring predictions given counterfactual values time-series data, individual observations can edited keeping observations original fitted values. type type prediction required. default scale linear predictors; alternative \"response\" scale response variable. Thus Poisson-distributed variable default predictions log-intensity type = \"response\" gives predicted intensity. ... used","code":""},{"path":"/reference/predict.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"predictions using dsem — predict.dsem","text":"matrix predicted values dimensions order corresponding argument newdata provided, tsdata . Predictions provided either link response scale, generated re-optimizing random effects condition MLE fixed effects, given new data.","code":""},{"path":"/reference/print.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Print fitted dsem object — print.dsem","title":"Print fitted dsem object — print.dsem","text":"Prints output fitted dsem model","code":""},{"path":"/reference/print.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print fitted dsem object — print.dsem","text":"","code":"# S3 method for dsem print(x, ...)"},{"path":"/reference/print.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print fitted dsem object — print.dsem","text":"x Output dsem ... used","code":""},{"path":"/reference/print.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Print fitted dsem object — print.dsem","text":"return value, called provide clean terminal output calling fitted object terminal.","code":""},{"path":"/reference/residuals.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate residuals — residuals.dsem","title":"Calculate residuals — residuals.dsem","text":"Calculate deviance response residuals dsem","code":""},{"path":"/reference/residuals.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate residuals — residuals.dsem","text":"","code":"# S3 method for dsem residuals(object, type = c(\"deviance\", \"response\"), ...)"},{"path":"/reference/residuals.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate residuals — residuals.dsem","text":"object Output dsem type type residuals compute (option \"deviance\" \"response\" now) ... used","code":""},{"path":"/reference/residuals.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Calculate residuals — residuals.dsem","text":"matrix residuals, order dimensions argument tsdata passed dsem.","code":""},{"path":"/reference/sea_otter.html","id":null,"dir":"Reference","previous_headings":"","what":"Sea otter trophic cascade — sea_otter","title":"Sea otter trophic cascade — sea_otter","text":"Data used demonstrate test trophic cascades options","code":""},{"path":"/reference/sea_otter.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sea otter trophic cascade — sea_otter","text":"","code":"data(sea_otter)"},{"path":"/reference/simulate.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate dsem — simulate.dsem","title":"Simulate dsem — simulate.dsem","text":"Simulate fitted dsem model","code":""},{"path":"/reference/simulate.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate dsem — simulate.dsem","text":"","code":"# S3 method for dsem simulate( object, nsim = 1, seed = NULL, variance = c(\"none\", \"random\", \"both\"), resimulate_gmrf = FALSE, ... )"},{"path":"/reference/simulate.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate dsem — simulate.dsem","text":"object Output dsem nsim number simulated data sets seed random seed variance whether ignore uncertainty fixed random effects, include estimation uncertainty random effects, include estimation uncertainty fixed random effects resimulate_gmrf whether resimulate GMRF based estimated simulated random effects (determined argument variance) ... used","code":""},{"path":"/reference/simulate.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate dsem — simulate.dsem","text":"Simulated data, either obj$simulate obj compiled TMB object, first simulating new GMRF calling obj$simulate.","code":""},{"path":"/reference/simulate.dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate dsem — simulate.dsem","text":"function conducts parametric bootstrap, .e., simulates new data conditional upon estimated values fixed random effects. user can optionally simulate new random effects conditional upon estimated covariance, simulate new fixed random effects conditional upon imprecision. Note simulate effect states x_tj measurement measurements fitted using family=\"fixed\", unless resimulate_gmrf=TRUE. latter case, GMRF resimulated given estimated path coefficients","code":""},{"path":"/reference/summary.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"summarize dsem — summary.dsem","title":"summarize dsem — summary.dsem","text":"summarize parameters fitted dynamic structural equation model","code":""},{"path":"/reference/summary.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"summarize dsem — summary.dsem","text":"","code":"# S3 method for dsem summary(object, ...)"},{"path":"/reference/summary.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"summarize dsem — summary.dsem","text":"object Output dsem ... used","code":""},{"path":"/reference/summary.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"summarize dsem — summary.dsem","text":"Returns data.frame summarizing estimated path coefficients, containing columns: path parsed path coefficient lag lag, e.g. 1 means predictor time t effects response time t+1 name Parameter name start Start value supplied, NA otherwise parameter Parameter number first Variable path treated predictor second Variable path treated response direction Whether path one-headed two-headed Estimate Maximum likelihood estimate Std_Error Estimated standard error Hessian matrix z_value Estimate divided Std_Error p_value P-value associated z_value using two-sided Wald test","code":""},{"path":"/reference/summary.dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"summarize dsem — summary.dsem","text":"DSEM specified using \"arrow lag\" notation, specifies set path coefficients exogenous variance parameters estimated. Function dsem estimates maximum likelihood value coefficients parameters maximizing log-marginal likelihood. Standard errors parameters calculated matrix second derivatives log-marginal likelihood (\"Hessian matrix\"). However, many users want associate individual parameters standard errors path coefficients specified using \"arrow lag\" notation. task complicated models path coefficients variance parameters specified share single value priori, assigned name NA hence assumed fixed value priori (coefficients parameters assigned value standard error). summary function therefore compiles MLE coefficients (including duplicating values path coefficients assigned value) standard error estimates, outputs table associates user-supplied path parameter names. also outputs z-score p-value arising two-sided Wald test (.e. comparing estimate divided standard error standard normal distribution).","code":""},{"path":"/reference/vcov.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract Variance-Covariance Matrix — vcov.dsem","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"extract covariance fixed effects, fixed random effects.","code":""},{"path":"/reference/vcov.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"","code":"# S3 method for dsem vcov(object, which = c(\"fixed\", \"random\", \"both\"), ...)"},{"path":"/reference/vcov.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"object output dsem whether extract covariance among fixed effects, random effects, ... ignored, method compatibility","code":""},{"path":"/reference/vcov.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"square matrix containing estimated covariances among parameter estimates model. dimensions dependend upon argument , determine whether fixed, random effects, outputted.","code":""},{"path":"/news/index.html","id":"dsem-101","dir":"Changelog","previous_headings":"","what":"dsem 1.0.1","title":"dsem 1.0.1","text":"Fix bug simulate.dsem keep changing interface dsem Update CITATION indicate accepted paper","code":""},{"path":"/news/index.html","id":"dsem-100","dir":"Changelog","previous_headings":"","what":"dsem 1.0.0","title":"dsem 1.0.0","text":"CRAN release: 2023-12-08 Initial public release","code":""}] +[{"path":"/articles/vignette.html","id":"comparison-with-dynamic-linear-models","dir":"Articles","previous_headings":"","what":"Comparison with dynamic linear models","title":"Demonstration of selected features","text":"first demonstrate dsem gives identical results dynlm well-known econometric model, Klein-1 model. Results show packages provide (almost) identical estimates standard errors. can also compare results using Laplace approximation obtained via numerical integration random effects using MCMC. example, MCMC results somewhat higher estimates exogenous variance parameters (presumably follow chi-squared distribution positive skewness), otherwise two produce similar estimates.","code":"data(KleinI, package=\"AER\") TS = ts(data.frame(KleinI, \"time\"=time(KleinI) - 1931)) # dynlm fm_cons <- dynlm(consumption ~ cprofits + L(cprofits) + I(pwage + gwage), data = TS) fm_inv <- dynlm(invest ~ cprofits + L(cprofits) + capital, data = TS) # fm_pwage <- dynlm(pwage ~ gnp + L(gnp) + time, data = TS) # dsem sem = \" # Link, lag, param_name cprofits -> consumption, 0, a1 cprofits -> consumption, 1, a2 pwage -> consumption, 0, a3 gwage -> consumption, 0, a3 cprofits -> invest, 0, b1 cprofits -> invest, 1, b2 capital -> invest, 0, b3 gnp -> pwage, 0, c2 gnp -> pwage, 1, c3 time -> pwage, 0, c1 \" tsdata = TS[,c(\"time\",\"gnp\",\"pwage\",\"cprofits\",'consumption', \"gwage\",\"invest\",\"capital\")] fit = dsem( sem=sem, tsdata = tsdata, newtonsteps = 0, estimate_delta0 = TRUE, control = dsem_control(quiet = TRUE) ) #> 1 regions found. #> Using 1 threads #> 1 regions found. #> Using 1 threads # Compile m1 = rbind( summary(fm_cons)$coef[-1,], summary(fm_inv)$coef[-1,], summary(fm_pwage)$coef[-1,] )[,1:2] m2 = summary(fit$opt$SD)[1:9,] m = rbind( data.frame(\"var\"=rownames(m1),m1,\"method\"=\"OLS\",\"eq\"=rep(c(\"C\",\"I\",\"Wp\"),each=3)), data.frame(\"var\"=rownames(m1),m2,\"method\"=\"GMRF\",\"eq\"=rep(c(\"C\",\"I\",\"Wp\"),each=3)) ) m = cbind(m, \"lower\"=m$Estimate-m$Std..Error, \"upper\"=m$Estimate+m$Std..Error ) # ggplot estimates longform = melt( as.data.frame(KleinI) ) longform$year = rep( time(KleinI), 9 ) p1 = ggplot( data=longform, aes(x=year, y=value) ) + facet_grid( rows=vars(variable), scales=\"free\" ) + geom_line( ) p2 = ggplot(data=m, aes(x=interaction(var,eq), y=Estimate, color=method)) + geom_point( position=position_dodge(0.9) ) + geom_errorbar( aes(ymax=as.numeric(upper),ymin=as.numeric(lower)), width=0.25, position=position_dodge(0.9)) # p3 = plot( as_fitted_DAG(fit) ) + expand_limits(x = c(-0.2,1) ) p4 = plot( as_fitted_DAG(fit, lag=1), text_size=4 ) + expand_limits(x = c(-0.2,1), y = c(-0.2,0) ) p1 p2 grid.arrange( arrangeGrob(p3, p4, nrow=2) ) library(tmbstan) # MCMC for both fixed and random effects mcmc = tmbstan( fit$obj, init=\"last.par.best\" ) summary_mcmc = summary(mcmc) # long-form data frame m1 = summary_mcmc$summary[1:17,c('mean','sd')] rownames(m1) = paste0( \"b\", seq_len(nrow(m1)) ) m2 = summary(fit$opt$SD)[1:17,c('Estimate','Std. Error')] m = rbind( data.frame('mean'=m1[,1], 'sd'=m1[,2], 'par'=rownames(m1), \"method\"=\"MCMC\"), data.frame('mean'=m2[,1], 'sd'=m2[,2], 'par'=rownames(m1), \"method\"=\"LA\") ) m$lower = m$mean - m$sd m$upper = m$mean + m$sd # plot ggplot(data=m, aes(x=par, y=mean, col=method)) + geom_point( position=position_dodge(0.9) ) + geom_errorbar( aes(ymax=as.numeric(upper),ymin=as.numeric(lower)), width=0.25, position=position_dodge(0.9)) #"},{"path":"/articles/vignette.html","id":"comparison-with-vector-autoregressive-models","dir":"Articles","previous_headings":"","what":"Comparison with vector autoregressive models","title":"Demonstration of selected features","text":"next demonstrate dsem gives similar results vector autoregressive (VAR) model. , analyze population abundance wolf moose populations Isle Royale 1959 2019, downloaded website (Vucetich, JA Peterson RO. 2012. population biology Isle Royale wolves moose: overview. URL: www.isleroyalewolf.org). dataset previously analyzed Chapter 14 User Manual R-package MARSS (Holmes, E. E., M. D. Scheuerell, E. J. Ward (2023) Analysis multivariate time-series using MARSS package. Version 3.11.8. NOAA Fisheries, Northwest Fisheries Science Center, 2725 Montlake Blvd E., Seattle, WA 98112, DOI: 10.5281/zenodo.5781847). , compare fits using dsem dynlm, well vector autoregressive model package vars, finally MARSS. Results show dsem can estimate parameters vector autoregressive model (VAM), exactly matches results vars, using dynlm, using MARSS.","code":"data(isle_royale) data = ts( log(isle_royale[,2:3]), start=1959) sem = \" # Link, lag, param_name wolves -> wolves, 1, arW moose -> wolves, 1, MtoW wolves -> moose, 1, WtoM moose -> moose, 1, arM \" # initial first without delta0 (to improve starting values) fit0 = dsem( sem = sem, tsdata = data, estimate_delta0 = FALSE, control = dsem_control(quiet=TRUE), getsd=FALSE ) #> Warning in nlminb(start = startpar, objective = fn, gradient = gr, control = #> nlminb.control, : NA/NaN function evaluation # Refit with delta0 fit = dsem( sem = sem, tsdata = data, estimate_delta0 = TRUE, control = dsem_control( quiet=TRUE, parameters = fit0$obj$env$parList()) ) # dynlm fm_wolf = dynlm( wolves ~ 1 + L(wolves) + L(moose), data=data ) # fm_moose = dynlm( moose ~ 1 + L(wolves) + L(moose), data=data ) # # MARSS library(MARSS) z.royale.dat <- t(scale(data.frame(data),center=TRUE,scale=FALSE)) royale.model.1 <- list( Z = \"identity\", B = \"unconstrained\", Q = \"diagonal and unequal\", R = \"zero\", U = \"zero\" ) kem.1 <- MARSS(z.royale.dat, model = royale.model.1) #> Success! algorithm run for 15 iterations. abstol and log-log tests passed. #> Alert: conv.test.slope.tol is 0.5. #> Test with smaller values (<0.1) to ensure convergence. #> #> MARSS fit is #> Estimation method: kem #> Convergence test: conv.test.slope.tol = 0.5, abstol = 0.001 #> Algorithm ran 15 (=minit) iterations and convergence was reached. #> Log-likelihood: -3.21765 #> AIC: 22.4353 AICc: 23.70964 #> #> Estimate #> B.(1,1) 0.8669 #> B.(2,1) -0.1240 #> B.(1,2) 0.1417 #> B.(2,2) 0.8176 #> Q.(X.wolves,X.wolves) 0.1341 #> Q.(X.moose,X.moose) 0.0284 #> x0.X.wolves 0.2324 #> x0.X.moose -0.6476 #> Initial states (x0) defined at t=0 #> #> Standard errors have not been calculated. #> Use MARSSparamCIs to compute CIs and bias estimates. SE <- MARSSparamCIs( kem.1 ) # Using VAR package library(vars) var = VAR( data, type=\"const\" ) ### Compile m1 = rbind( summary(fm_wolf)$coef[-1,], summary(fm_moose)$coef[-1,] )[,1:2] m2 = summary(fit$opt$SD)[1:4,] #m2 = cbind( \"Estimate\"=fit$opt$par, \"Std. Error\"=fit$opt$SD$par.fixed )[1:4,] m3 = cbind( SE$parMean[c(1,3,2,4)], SE$par.se$B[c(1,3,2,4)] ) colnames(m3) = colnames(m2) m4 = rbind( summary(var$varresult$wolves)$coef[-3,], summary(var$varresult$moose)$coef[-3,] )[,1:2] # Bundle m = rbind( data.frame(\"var\"=rownames(m1), m1, \"method\"=\"dynlm\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)), data.frame(\"var\"=rownames(m1), m2, \"method\"=\"dsem\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)), data.frame(\"var\"=rownames(m1), m3, \"method\"=\"MARSS\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)), data.frame(\"var\"=rownames(m1), m4, \"method\"=\"vars\", \"eq\"=rep(c(\"Wolf\",\"Moose\"),each=2)) ) #knitr::kable( m1, digits=3) #knitr::kable( m2, digits=3) m = cbind(m, \"lower\"=m$Estimate-m$Std..Error, \"upper\"=m$Estimate+m$Std..Error ) # ggplot estimates ... interaction(x,y) causes an error sometimes library(ggplot2) library(ggpubr) library(ggraph) longform = reshape( isle_royale, idvar = \"year\", direction=\"long\", varying=list(2:3), v.names=\"abundance\", timevar=\"species\", times=c(\"wolves\",\"moose\") ) p1 = ggplot( data=longform, aes(x=year, y=abundance) ) + facet_grid( rows=vars(species), scales=\"free\" ) + geom_point( ) p2 = ggplot(data=m, aes(x=interaction(var,eq), y=Estimate, color=method)) + geom_point( position=position_dodge(0.9) ) + geom_errorbar( aes(ymax=as.numeric(upper),ymin=as.numeric(lower)), width=0.25, position=position_dodge(0.9)) # p3 = plot( as_fitted_DAG(fit, lag=1), rotation=0 ) + geom_edge_loop( aes( label=round(weight,2), direction=0)) + #arrow=arrow(), , angle_calc=\"along\", label_dodge=grid::unit(10,\"points\") ) expand_limits(x = c(-0.1,0) ) ggarrange( p1, p2, p3, labels = c(\"Time-series data\", \"Estimated effects\", \"Fitted path digram\"), ncol = 1, nrow = 3)"},{"path":"/articles/vignette.html","id":"multi-causal-ecosystem-synthesis","dir":"Articles","previous_headings":"","what":"Multi-causal ecosystem synthesis","title":"Demonstration of selected features","text":"next replicate analysis involving climate, forage fishes, stomach contents, recruitment predatory fish. results discussed paper describing dsem.","code":"data(bering_sea) Z = ts( bering_sea ) family = rep('fixed', ncol(bering_sea)) # Specify model sem = \" # Link, lag, param_name log_seaice -> log_CP, 0, seaice_to_CP log_CP -> log_Cfall, 0, CP_to_Cfall log_CP -> log_Esummer, 0, CP_to_E log_PercentEuph -> log_RperS, 0, Seuph_to_RperS log_PercentCop -> log_RperS, 0, Scop_to_RperS log_Esummer -> log_PercentEuph, 0, Esummer_to_Suph log_Cfall -> log_PercentCop, 0, Cfall_to_Scop log_SSB -> log_RperS, 0, SSB_to_RperS log_seaice -> log_seaice, 1, AR1, 0.001 log_CP -> log_CP, 1, AR2, 0.001 log_Cspring -> log_Cspring, 1, AR3, 0.001 log_Cfall -> log_Cfall, 1, AR4, 0.001 log_Esummer -> log_Esummer, 1, AR5, 0.001 log_SSB -> log_SSB, 1, AR6, 0.001 log_RperS -> log_RperS, 1, AR7, 0.001 log_PercentEuph -> log_PercentEuph, 1, AR8, 0.001 log_PercentCop -> log_PercentCop, 1, AR9, 0.001 \" # Fit fit = dsem( sem = sem, tsdata = Z, family = family, control = dsem_control(use_REML=FALSE, quiet=TRUE) ) #> Warning in nlminb(start = startpar, objective = fn, gradient = gr, control = #> nlminb.control, : NA/NaN function evaluation #> Warning in nlminb(start = startpar, objective = fn, gradient = gr, control = #> nlminb.control, : NA/NaN function evaluation ParHat = fit$obj$env$parList() # summary( fit ) # Timeseries plot oldpar <- par(no.readonly = TRUE) par( mfcol=c(3,3), mar=c(2,2,2,0), mgp=c(2,0.5,0), tck=-0.02 ) for(i in 1:ncol(bering_sea)){ tmp = bering_sea[,i,drop=FALSE] tmp = cbind( tmp, \"PSEM\"=ParHat$x_tj[,i] ) SD = as.list(fit$opt$SD,what=\"Std.\")$x_tj[,i] tmp = cbind( tmp, outer(tmp[,2],c(1,1)) + outer(ifelse(is.na(SD),0,SD),c(-1,1)) ) # plot( x=rownames(bering_sea), y=tmp[,1], ylim=range(tmp,na.rm=TRUE), type=\"p\", main=colnames(bering_sea)[i], pch=20, cex=2 ) lines( x=rownames(bering_sea), y=tmp[,2], type=\"l\", lwd=2, col=\"blue\", lty=\"solid\" ) polygon( x=c(rownames(bering_sea),rev(rownames(bering_sea))), y=c(tmp[,3],rev(tmp[,4])), col=rgb(0,0,1,0.2), border=NA ) } par(oldpar) # library(phylopath) library(ggplot2) library(ggpubr) library(reshape) library(gridExtra) longform = melt( bering_sea ) longform$year = rep( 1963:2023, ncol(bering_sea) ) p0 = ggplot( data=longform, aes(x=year, y=value) ) + facet_grid( rows=vars(variable), scales=\"free\" ) + geom_point( ) p1 = plot( (as_fitted_DAG(fit)), edge.width=1, type=\"width\", text_size=4, show.legend=FALSE, arrow = grid::arrow(type='closed', 18, grid::unit(10,'points')) ) + scale_x_continuous(expand = c(0.4, 0.1)) p1$layers[[1]]$mapping$edge_width = 1 p2 = plot( (as_fitted_DAG(fit, what=\"p_value\")), edge.width=1, type=\"width\", text_size=4, show.legend=FALSE, colors=c('black', 'black'), arrow = grid::arrow(type='closed', 18, grid::unit(10,'points')) ) + scale_x_continuous(expand = c(0.4, 0.1)) p2$layers[[1]]$mapping$edge_width = 0.5 #grid.arrange( arrangeGrob( p0+ggtitle(\"timeseries\"), # arrangeGrob( p1+ggtitle(\"Estimated path diagram\"), # p2+ggtitle(\"Estimated p-values\"), nrow=2), ncol=2 ) ) ggarrange(p1, p2, labels = c(\"Simultaneous effects\", \"Two-sided p-value\"), ncol = 1, nrow = 2)"},{"path":"/articles/vignette.html","id":"site-replicated-trophic-cascade","dir":"Articles","previous_headings":"","what":"Site-replicated trophic cascade","title":"Demonstration of selected features","text":"Finally, replicate analysis involving trophic cascade involving sea stars predators, sea urchin consumers, kelp producers. , results discussed paper describing dsem.","code":"data(sea_otter) Z = ts( sea_otter[,-1] ) # Specify model sem = \" Pycno_CANNERY_DC -> log_Urchins_CANNERY_DC, 0, x2 log_Urchins_CANNERY_DC -> log_Kelp_CANNERY_DC, 0, x3 log_Otter_Count_CANNERY_DC -> log_Kelp_CANNERY_DC, 0, x4 Pycno_CANNERY_UC -> log_Urchins_CANNERY_UC, 0, x2 log_Urchins_CANNERY_UC -> log_Kelp_CANNERY_UC, 0, x3 log_Otter_Count_CANNERY_UC -> log_Kelp_CANNERY_UC, 0, x4 Pycno_HOPKINS_DC -> log_Urchins_HOPKINS_DC, 0, x2 log_Urchins_HOPKINS_DC -> log_Kelp_HOPKINS_DC, 0, x3 log_Otter_Count_HOPKINS_DC -> log_Kelp_HOPKINS_DC, 0, x4 Pycno_HOPKINS_UC -> log_Urchins_HOPKINS_UC, 0, x2 log_Urchins_HOPKINS_UC -> log_Kelp_HOPKINS_UC, 0, x3 log_Otter_Count_HOPKINS_UC -> log_Kelp_HOPKINS_UC, 0, x4 Pycno_LOVERS_DC -> log_Urchins_LOVERS_DC, 0, x2 log_Urchins_LOVERS_DC -> log_Kelp_LOVERS_DC, 0, x3 log_Otter_Count_LOVERS_DC -> log_Kelp_LOVERS_DC, 0, x4 Pycno_LOVERS_UC -> log_Urchins_LOVERS_UC, 0, x2 log_Urchins_LOVERS_UC -> log_Kelp_LOVERS_UC, 0, x3 log_Otter_Count_LOVERS_UC -> log_Kelp_LOVERS_UC, 0, x4 Pycno_MACABEE_DC -> log_Urchins_MACABEE_DC, 0, x2 log_Urchins_MACABEE_DC -> log_Kelp_MACABEE_DC, 0, x3 log_Otter_Count_MACABEE_DC -> log_Kelp_MACABEE_DC, 0, x4 Pycno_MACABEE_UC -> log_Urchins_MACABEE_UC, 0, x2 log_Urchins_MACABEE_UC -> log_Kelp_MACABEE_UC, 0, x3 log_Otter_Count_MACABEE_UC -> log_Kelp_MACABEE_UC, 0, x4 Pycno_OTTER_PT_DC -> log_Urchins_OTTER_PT_DC, 0, x2 log_Urchins_OTTER_PT_DC -> log_Kelp_OTTER_PT_DC, 0, x3 log_Otter_Count_OTTER_PT_DC -> log_Kelp_OTTER_PT_DC, 0, x4 Pycno_OTTER_PT_UC -> log_Urchins_OTTER_PT_UC, 0, x2 log_Urchins_OTTER_PT_UC -> log_Kelp_OTTER_PT_UC, 0, x3 log_Otter_Count_OTTER_PT_UC -> log_Kelp_OTTER_PT_UC, 0, x4 Pycno_PINOS_CEN -> log_Urchins_PINOS_CEN, 0, x2 log_Urchins_PINOS_CEN -> log_Kelp_PINOS_CEN, 0, x3 log_Otter_Count_PINOS_CEN -> log_Kelp_PINOS_CEN, 0, x4 Pycno_SIREN_CEN -> log_Urchins_SIREN_CEN, 0, x2 log_Urchins_SIREN_CEN -> log_Kelp_SIREN_CEN, 0, x3 log_Otter_Count_SIREN_CEN -> log_Kelp_SIREN_CEN, 0, x4 # AR1 Pycno_CANNERY_DC -> Pycno_CANNERY_DC, 1, ar1 log_Urchins_CANNERY_DC -> log_Urchins_CANNERY_DC, 1, ar2 log_Otter_Count_CANNERY_DC -> log_Otter_Count_CANNERY_DC, 1, ar3 log_Kelp_CANNERY_DC -> log_Kelp_CANNERY_DC, 1, ar4 Pycno_CANNERY_UC -> Pycno_CANNERY_UC, 1, ar1 log_Urchins_CANNERY_UC -> log_Urchins_CANNERY_UC, 1, ar2 log_Otter_Count_CANNERY_UC -> log_Otter_Count_CANNERY_UC, 1, ar3 log_Kelp_CANNERY_UC -> log_Kelp_CANNERY_UC, 1, ar4 Pycno_HOPKINS_DC -> Pycno_HOPKINS_DC, 1, ar1 log_Urchins_HOPKINS_DC -> log_Urchins_HOPKINS_DC, 1, ar2 log_Otter_Count_HOPKINS_DC -> log_Otter_Count_HOPKINS_DC, 1, ar3 log_Kelp_HOPKINS_DC -> log_Kelp_HOPKINS_DC, 1, ar4 Pycno_HOPKINS_UC -> Pycno_HOPKINS_UC, 1, ar1 log_Urchins_HOPKINS_UC -> log_Urchins_HOPKINS_UC, 1, ar2 log_Otter_Count_HOPKINS_UC -> log_Otter_Count_HOPKINS_UC, 1, ar3 log_Kelp_HOPKINS_UC -> log_Kelp_HOPKINS_UC, 1, ar4 Pycno_LOVERS_DC -> Pycno_LOVERS_DC, 1, ar1 log_Urchins_LOVERS_DC -> log_Urchins_LOVERS_DC, 1, ar2 log_Otter_Count_LOVERS_DC -> log_Otter_Count_LOVERS_DC, 1, ar3 log_Kelp_LOVERS_DC -> log_Kelp_LOVERS_DC, 1, ar4 Pycno_LOVERS_UC -> Pycno_LOVERS_UC, 1, ar1 log_Urchins_LOVERS_UC -> log_Urchins_LOVERS_UC, 1, ar2 log_Otter_Count_LOVERS_UC -> log_Otter_Count_LOVERS_UC, 1, ar3 log_Kelp_LOVERS_UC -> log_Kelp_LOVERS_UC, 1, ar4 Pycno_MACABEE_DC -> Pycno_MACABEE_DC, 1, ar1 log_Urchins_MACABEE_DC -> log_Urchins_MACABEE_DC, 1, ar2 log_Otter_Count_MACABEE_DC -> log_Otter_Count_MACABEE_DC, 1, ar3 log_Kelp_MACABEE_DC -> log_Kelp_MACABEE_DC, 1, ar4 Pycno_MACABEE_UC -> Pycno_MACABEE_UC, 1, ar1 log_Urchins_MACABEE_UC -> log_Urchins_MACABEE_UC, 1, ar2 log_Otter_Count_MACABEE_UC -> log_Otter_Count_MACABEE_UC, 1, ar3 log_Kelp_MACABEE_UC -> log_Kelp_MACABEE_UC, 1, ar4 Pycno_OTTER_PT_DC -> Pycno_OTTER_PT_DC, 1, ar1 log_Urchins_OTTER_PT_DC -> log_Urchins_OTTER_PT_DC, 1, ar2 log_Otter_Count_OTTER_PT_DC -> log_Otter_Count_OTTER_PT_DC, 1, ar3 log_Kelp_OTTER_PT_DC -> log_Kelp_OTTER_PT_DC, 1, ar4 Pycno_OTTER_PT_UC -> Pycno_OTTER_PT_UC, 1, ar1 log_Urchins_OTTER_PT_UC -> log_Urchins_OTTER_PT_UC, 1, ar2 log_Otter_Count_OTTER_PT_UC -> log_Otter_Count_OTTER_PT_UC, 1, ar3 log_Kelp_OTTER_PT_UC -> log_Kelp_OTTER_PT_UC, 1, ar4 Pycno_PINOS_CEN -> Pycno_PINOS_CEN, 1, ar1 log_Urchins_PINOS_CEN -> log_Urchins_PINOS_CEN, 1, ar2 log_Otter_Count_PINOS_CEN -> log_Otter_Count_PINOS_CEN, 1, ar3 log_Kelp_PINOS_CEN -> log_Kelp_PINOS_CEN, 1, ar4 Pycno_SIREN_CEN -> Pycno_SIREN_CEN, 1, ar1 log_Urchins_SIREN_CEN -> log_Urchins_SIREN_CEN, 1, ar2 log_Otter_Count_SIREN_CEN -> log_Otter_Count_SIREN_CEN, 1, ar3 log_Kelp_SIREN_CEN -> log_Kelp_SIREN_CEN, 1, ar4 \" # Fit model fit = dsem( sem = sem, tsdata = Z, control = dsem_control(use_REML=FALSE, quiet=TRUE) ) # summary( fit ) # library(phylopath) library(ggplot2) library(ggpubr) get_part = function(x){ vars = c(\"log_Kelp\",\"log_Otter\",\"log_Urchin\",\"Pycno\") index = sapply( vars, FUN=\\(y) grep(y,rownames(x$coef))[1] ) x$coef = x$coef[index,index] dimnames(x$coef) = list( vars, vars ) return(x) } p1 = plot( get_part(as_fitted_DAG(fit)), type=\"width\", show.legend=FALSE) p1$layers[[1]]$mapping$edge_width = 0.5 p2 = plot( get_part(as_fitted_DAG(fit, what=\"p_value\" )), type=\"width\", show.legend=FALSE, colors=c('black', 'black')) p2$layers[[1]]$mapping$edge_width = 0.1 longform = melt( sea_otter[,-1], as.is=TRUE ) #> Warning in type.convert.default(X[[i]], ...): 'as.is' should be specified by #> the caller; using TRUE longform$X1 = 1999:2019[longform$X1] #> Warning in 1999:2019[longform$X1]: numerical expression has 1008 elements: only #> the first used longform$Site = gsub( \"log_Kelp_\", \"\", gsub( \"log_Urchins_\", \"\", gsub( \"Pycno_\", \"\", gsub( \"log_Otter_Count_\", \"\", longform$X2)))) longform$Species = sapply( seq_len(nrow(longform)), FUN=\\(i)gsub(longform$Site[i],\"\",longform$X2[i]) ) p3 = ggplot( data=longform, aes(x=X1, y=value, col=Species) ) + facet_grid( rows=vars(Site), scales=\"free\" ) + geom_line( ) ggarrange(p1 + scale_x_continuous(expand = c(0.3, 0)), p2 + scale_x_continuous(expand = c(0.3, 0)), labels = c(\"Simultaneous effects\", \"Two-sided p-value\"), ncol = 1, nrow = 2)"},{"path":"/authors.html","id":null,"dir":"","previous_headings":"","what":"Authors","title":"Authors and Citation","text":"James Thorson. Author, maintainer.","code":""},{"path":"/authors.html","id":"citation","dir":"","previous_headings":"","what":"Citation","title":"Authors and Citation","text":"Thorson, J. T., Andrews, . G., Essington, T., & Large, S. (press). Dynamic structural equation models synthesize ecosystem dynamics constrained ecological mechanisms. Methods Ecology Evolution.","code":"@Article{, title = {Dynamic structural equation models synthesize ecosystem dynamics constrained by ecological mechanisms}, journal = {Methods in Ecology and Evolution}, year = {2024}, author = {J. T. Thorson and A. G. Andrews and T. Essington and S. Large}, }"},{"path":"/index.html","id":"dynamic-structural-equation-models","dir":"","previous_headings":"","what":"Dynamic structural equation models","title":"Fit Dynamic Structural Equation Models","text":"Package dsem fits dynamic structural equation models, includes nested submodels: structural equation models vector autoregressive models dynamic factor analysis state-space autoregressive integrated moving average (ARIMA) models model several advantages: estimates direct, indirect, total effects among system variables, including simultaneous lagged effects recursive (cyclic) dependencies can estimate cumulative outcome press pulse experiments initial conditions differ stationary distribution system dynamics estimates structural linkages regression slopes jointly imputing missing values /measurement errors rapidly fitted Gaussian Markov random field (GMRF) Generalized Linear Mixed Model (GLMM), speed asymptotics associated allows granular control number parameters (restrictions parameters) used structure covariance among variables time, phylosem specifically intended minimal implementation, uses standard packages simplify input/output formatting: Input: time-series defined using class ts, NA missing values Input: structural trade-offs specified using syntax defined package sem Output: visualizing estimated trade-offs using igraph Output: access model output using standard S3-generic functions including summary, predict, residuals, simulate, AIC Please see package vignettes details regarding syntax features.","code":""},{"path":"/reference/TMBAIC.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate marginal AIC for a fitted model — TMBAIC","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"TMBAIC calculates AIC given model fit","code":""},{"path":"/reference/TMBAIC.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"","code":"TMBAIC(opt, k = 2, n = Inf)"},{"path":"/reference/TMBAIC.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"opt output nlminb optim k penalty additional fixed effects (default=2, AIC) n sample size, use AICc calculation (default=Inf, AICc=AIC)","code":""},{"path":"/reference/TMBAIC.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Calculate marginal AIC for a fitted model — TMBAIC","text":"AIC, parsimonious model AIC relative candidate models","code":""},{"path":"/reference/as_fitted_DAG.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert output from package dsem to phylopath — as_fitted_DAG","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"Convert dsem phylopath output","code":""},{"path":"/reference/as_fitted_DAG.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"","code":"as_fitted_DAG(fit, lag = 0, what = \"Estimate\", direction = 1)"},{"path":"/reference/as_fitted_DAG.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"fit Output dsem lag lag output whether output estimates =\"Estimate\", standard errors =\"Std_Error\" p-values =\"Std_Error\" direction whether include one-sided arrows direction=1, one- two-sided arrows direction=c(1,2)","code":""},{"path":"/reference/as_fitted_DAG.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Convert output from package dsem to phylopath — as_fitted_DAG","text":"Convert output format supplied est_DAG","code":""},{"path":"/reference/as_sem.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert dsem to sem output — as_sem","title":"Convert dsem to sem output — as_sem","text":"Convert output package dsem sem","code":""},{"path":"/reference/as_sem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert dsem to sem output — as_sem","text":"","code":"as_sem(object, lag = 0)"},{"path":"/reference/as_sem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert dsem to sem output — as_sem","text":"object Output dsem lag lag extract visualize","code":""},{"path":"/reference/as_sem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Convert dsem to sem output — as_sem","text":"Convert output format supplied sem","code":""},{"path":"/reference/bering_sea.html","id":null,"dir":"Reference","previous_headings":"","what":"Bering Sea marine ecosystem — bering_sea","title":"Bering Sea marine ecosystem — bering_sea","text":"Data used demonstrate test ecosystem synthesis","code":""},{"path":"/reference/bering_sea.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Bering Sea marine ecosystem — bering_sea","text":"","code":"data(bering_sea)"},{"path":"/reference/classify_variables.html","id":null,"dir":"Reference","previous_headings":"","what":"Classify variables path — classify_variables","title":"Classify variables path — classify_variables","text":"classify_variables copied sem:::classifyVariables","code":""},{"path":"/reference/classify_variables.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Classify variables path — classify_variables","text":"","code":"classify_variables(model)"},{"path":"/reference/classify_variables.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Classify variables path — classify_variables","text":"model SEM model","code":""},{"path":"/reference/classify_variables.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Classify variables path — classify_variables","text":"Tagged-list defining exogenous endogenous variables","code":""},{"path":"/reference/classify_variables.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Classify variables path — classify_variables","text":"Copied package `sem` licence GPL (>= 2) permission John Fox","code":""},{"path":"/reference/dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit dynamic structural equation model — dsem","title":"Fit dynamic structural equation model — dsem","text":"Fits dynamic structural equation model","code":""},{"path":"/reference/dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit dynamic structural equation model — dsem","text":"","code":"dsem( sem, tsdata, family = rep(\"fixed\", ncol(tsdata)), estimate_delta0 = FALSE, control = dsem_control(), ... )"},{"path":"/reference/dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit dynamic structural equation model — dsem","text":"sem Specification time-series structural equation model structure including lagged simultaneous effects. See Details section make_dsem_ram description tsdata time-series data, outputted using ts family Character-vector listing distribution used column tsdata, element must fixed normal. family=\"fixed\" default behavior assumes given variable measured exactly. options correspond different specifications measurement error. estimate_delta0 Boolean indicating whether estimate deviations equilibrium initial year fixed effects, alternatively assume dynamics start stochastic draw away stationary distribution control Output dsem_control, used define user settings, see documentation function details. ... Additional parameters passed fit_tmb","code":""},{"path":"/reference/dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit dynamic structural equation model — dsem","text":"object (list) class `dsem`. Elements include: obj TMB object MakeADFun ram RAM parsed make_dsem_ram model SEM structure parsed make_dsem_ram intermediate description model linkages tmb_inputs list inputs passed MakeADFun opt output fit_tmb","code":""},{"path":"/reference/dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit dynamic structural equation model — dsem","text":"DSEM involves (minimum): Time series matrix \\(\\mathbf X\\) column \\(\\mathbf x_c\\) variable c time-series; Path diagram user-supplied specification path coefficients, define precision (inverse covariance) \\(\\mathbf Q\\) matrix state-variables see make_dsem_ram details math involved. model also estimates time-series mean \\( \\mathbf{\\mu}_c \\) variable. mean precision matrix therefore define Gaussian Markov random field \\(\\mathbf X\\): $$ \\mathrm{vec}(\\mathbf X) \\sim \\mathrm{MVN}( \\mathrm{vec}(\\mathbf{I_T} \\otimes \\mathbf{\\mu}), \\mathbf{Q}^{-1}) $$ Users can specify distribution measurement errors (assume variables measured without error) using argument family. defines link-function \\(g_c(.)\\) distribution \\(f_c(.)\\) time-series \\(c\\): $$ y_{t,c} \\sim f_c( g_c^{-1}( x_{t,c} ), \\theta_c )$$ dsem estimates specified coefficients, time-series means \\(\\mu_c\\), distribution measurement errors \\(\\theta_c\\) via maximizing log-marginal likelihood, also estimating state-variables \\(x_{t,c}\\). summary.dsem assembles estimates standard errors easy--read format. Standard errors fixed effects (path coefficients, exogenoux variance parameters, measurement error parameters) estimated matrix second derivatives log-marginal likelihod, standard errors random effects (.e., missing state-space variables) estimated generalization method (see sdreport details).","code":""},{"path":"/reference/dsem.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Fit dynamic structural equation model — dsem","text":"**Introducing package, features, comparison software (cite using dsem):** Thorson, J. T., Andrews, ., Essington, T., Large, S. (review). Dynamic structural equation models synthesize ecosystem dynamics constrained ecological mechanisms.","code":""},{"path":"/reference/dsem.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Fit dynamic structural equation model — dsem","text":"","code":"# Define model sem = \" # Link, lag, param_name cprofits -> consumption, 0, a1 cprofits -> consumption, 1, a2 pwage -> consumption, 0, a3 gwage -> consumption, 0, a3 cprofits -> invest, 0, b1 cprofits -> invest, 1, b2 capital -> invest, 0, b3 gnp -> pwage, 0, c2 gnp -> pwage, 1, c3 time -> pwage, 0, c1 \" # Load data data(KleinI, package=\"AER\") TS = ts(data.frame(KleinI, \"time\"=time(KleinI) - 1931)) tsdata = TS[,c(\"time\",\"gnp\",\"pwage\",\"cprofits\",'consumption', \"gwage\",\"invest\",\"capital\")] # Fit model fit = dsem( sem=sem, tsdata = tsdata, newtonsteps = 0, estimate_delta0 = TRUE, control = dsem_control(quiet=TRUE) ) #> NOTE: adding 8 variances to the model #> 1 regions found. #> Using 1 threads #> 1 regions found. #> Using 1 threads summary( fit ) #> path lag name start parameter first #> 1 cprofits -> consumption 0 a1 1 cprofits #> 2 cprofits -> consumption 1 a2 2 cprofits #> 3 pwage -> consumption 0 a3 3 pwage #> 4 gwage -> consumption 0 a3 3 gwage #> 5 cprofits -> invest 0 b1 4 cprofits #> 6 cprofits -> invest 1 b2 5 cprofits #> 7 capital -> invest 0 b3 6 capital #> 8 gnp -> pwage 0 c2 7 gnp #> 9 gnp -> pwage 1 c3 8 gnp #> 10 time -> pwage 0 c1 9 time #> 11 cprofits <-> cprofits 0 V[cprofits] 10 cprofits #> 12 consumption <-> consumption 0 V[consumption] 11 consumption #> 13 pwage <-> pwage 0 V[pwage] 12 pwage #> 14 gwage <-> gwage 0 V[gwage] 13 gwage #> 15 invest <-> invest 0 V[invest] 14 invest #> 16 capital <-> capital 0 V[capital] 15 capital #> 17 gnp <-> gnp 0 V[gnp] 16 gnp #> 18 time <-> time 0 V[time] 17 time #> second direction Estimate Std_Error z_value p_value #> 1 consumption 1 0.19323180 0.08199227 2.356708 1.843776e-02 #> 2 consumption 1 0.08942121 0.08136332 1.099036 2.717524e-01 #> 3 consumption 1 0.79625660 0.03594117 22.154444 9.451951e-109 #> 4 consumption 1 0.79625660 0.03594117 22.154444 9.451951e-109 #> 5 invest 1 0.48138159 0.08740017 5.507788 3.633710e-08 #> 6 invest 1 0.33084594 0.09069259 3.647993 2.642966e-04 #> 7 invest 1 -0.11150749 0.02408258 -4.630214 3.652878e-06 #> 8 pwage 1 0.44041578 0.02921826 15.073305 2.426629e-51 #> 9 pwage 1 0.14476508 0.03370695 4.294813 1.748404e-05 #> 10 pwage 1 0.13029843 0.02882713 4.519994 6.184145e-06 #> 11 cprofits 2 4.10929066 0.63141042 6.508114 7.610041e-11 #> 12 consumption 2 0.92285209 0.14239888 6.480754 9.126542e-11 #> 13 pwage 2 0.69302971 0.10701074 6.476263 9.402191e-11 #> 14 gwage 2 1.90952952 0.29464657 6.480746 9.127028e-11 #> 15 invest 2 0.91015014 0.14046555 6.479525 9.201155e-11 #> 16 capital 2 9.67951927 1.49358531 6.480727 9.128145e-11 #> 17 gnp 2 10.32018719 1.58101797 6.527558 6.685044e-11 #> 18 time 2 6.05530028 0.93435296 6.480742 9.127276e-11 plot( fit ) plot( fit, edge_label=\"value\" )"},{"path":"/reference/dsem_control.html","id":null,"dir":"Reference","previous_headings":"","what":"Detailed control for dsem structure — dsem_control","title":"Detailed control for dsem structure — dsem_control","text":"Define list control parameters. Note format input likely change rapidly dsem","code":""},{"path":"/reference/dsem_control.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Detailed control for dsem structure — dsem_control","text":"","code":"dsem_control( quiet = FALSE, run_model = TRUE, use_REML = TRUE, parameters = NULL, map = NULL )"},{"path":"/reference/dsem_control.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Detailed control for dsem structure — dsem_control","text":"quiet Boolean indicating whether run model printing messages terminal ; run_model Boolean indicating whether estimate parameters (default), instead return model inputs compiled TMB object without running; use_REML Boolean indicating whether treat non-variance fixed effects random, either motigate bias estimated variance parameters improve efficiency parameter estimation given correlated fixed random effects parameters list fixed random effects, e.g., constructed dsem modified hand (helpful advanced users change starting values restart intended values) map list fixed mirrored parameters, constructed dsem default available override default pass MakeADFun","code":""},{"path":"/reference/dsem_control.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Detailed control for dsem structure — dsem_control","text":"S3 object class \"dsem_control\" specifies detailed model settings, allowing user specification also specifying default values","code":""},{"path":"/reference/fit_tmb.html","id":null,"dir":"Reference","previous_headings":"","what":"Optimize a TMB model — fit_tmb","title":"Optimize a TMB model — fit_tmb","text":"fit_tmb runs TMB model generates standard diagnostics","code":""},{"path":"/reference/fit_tmb.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Optimize a TMB model — fit_tmb","text":"","code":"fit_tmb( obj, fn = obj$fn, gr = obj$gr, startpar = NULL, lower = -Inf, upper = Inf, getsd = TRUE, control = list(eval.max = 10000, iter.max = 10000, trace = 1), bias.correct = FALSE, bias.correct.control = list(sd = FALSE, split = NULL, nsplit = NULL, vars_to_correct = NULL), savedir = NULL, loopnum = 2, newtonsteps = 0, n = Inf, getReportCovariance = FALSE, getJointPrecision = FALSE, getHessian = FALSE, quiet = FALSE, start_time_elapsed = as.difftime(\"0:0:0\"), ... )"},{"path":"/reference/fit_tmb.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Optimize a TMB model — fit_tmb","text":"obj compiled TMB object fn Function minimized. Must return scalar value. . gr Optional function takes arguments objective evaluates gradient objective first argument. Must return vector long startpar startpar Starting values fixed effects (default NULL uses obj$par) lower, upper vectors lower upper bounds, replicated long start. unspecified, parameters assumed unconstrained. getsd Boolean indicating whether run standard error calculation; see sdreport details control list control parameters. details see nlminb bias.correct Boolean indicating whether epsilon bias-correction; see sdreport fit_tmbfor details bias.correct.control tagged list options epsilon bias-correction, vars_to_correct character-vector ADREPORT variables bias-corrected savedir directory save results (savedir=NULL, results saved) loopnum number times re-start optimization (loopnum=3 sometimes achieves lower final gradient loopnum=1) newtonsteps Integer specifying number extra newton steps take optimization (alternative loopnum). newtonstep requires calculating Hessian matrix therefore slow. well-behaved models, Newton step typically decrease maximum gradient loglikelihood respect fixed effect, therefore option can used achieve arbitrarily low final gradient given sufficient time well-behaved models. However, option also perform strangely unexpected consequences poorly-behaved models, e.g., fixed effects upper lower bounds. n sample sizes (n!=Inf n used calculate BIC AICc) getReportCovariance Get full covariance matrix ADREPORTed variables? getJointPrecision Optional. Return full joint precision matrix random effects parameters? getHessian return Hessian usage later code quiet Boolean whether print additional messages results terminal start_time_elapsed much time elapsed prior calling fit_tmb, use, e.g., calling fit_tmb multiple times sequence, start_time_elapsed = opt_previous$time_for_run ... list settings pass sdreport","code":""},{"path":"/reference/fit_tmb.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Optimize a TMB model — fit_tmb","text":"standard output nlminb, except additional diagnostics timing info, new slot containing output sdreport","code":""},{"path":"/reference/fit_tmb.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Optimize a TMB model — fit_tmb","text":"details see doi:10.1016/j.fishres.2015.11.016","code":""},{"path":"/reference/isle_royale.html","id":null,"dir":"Reference","previous_headings":"","what":"Isle Royale wolf and moose — isle_royale","title":"Isle Royale wolf and moose — isle_royale","text":"Data used demonstrate test cross-lagged (vector autoregressive) models","code":""},{"path":"/reference/isle_royale.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Isle Royale wolf and moose — isle_royale","text":"","code":"data(isle_royale)"},{"path":"/reference/isle_royale.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Isle Royale wolf and moose — isle_royale","text":"Data extracted file \"Data_wolves_moose_Isle_Royale_June2019.csv\" available https://isleroyalewolf.org/data/data/home.html obtained 2023-06-23. Reproduced permission John Vucetich, generated Wolves Moose Isle Royale project.","code":""},{"path":"/reference/isle_royale.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Isle Royale wolf and moose — isle_royale","text":"Vucetich, JA Peterson RO. 2012. population biology Isle Royale wolves moose: overview. https://www.isleroyalewolf.org","code":""},{"path":"/reference/list_parameters.html","id":null,"dir":"Reference","previous_headings":"","what":"List fixed and random effects — list_parameters","title":"List fixed and random effects — list_parameters","text":"list_parameters lists fixed random effects","code":""},{"path":"/reference/list_parameters.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"List fixed and random effects — list_parameters","text":"","code":"list_parameters(Obj, verbose = TRUE)"},{"path":"/reference/list_parameters.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"List fixed and random effects — list_parameters","text":"Obj Compiled TMB object verbose Boolean, whether print messages terminal","code":""},{"path":"/reference/list_parameters.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"List fixed and random effects — list_parameters","text":"Tagged-list fixed random effects, returned invisibly printed screen","code":""},{"path":"/reference/logLik.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Marglinal log-likelihood — logLik.dsem","title":"Marglinal log-likelihood — logLik.dsem","text":"Extract (marginal) log-likelihood dsem model","code":""},{"path":"/reference/logLik.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Marglinal log-likelihood — logLik.dsem","text":"","code":"# S3 method for dsem logLik(object, ...)"},{"path":"/reference/logLik.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Marglinal log-likelihood — logLik.dsem","text":"object Output dsem ... used","code":""},{"path":"/reference/logLik.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Marglinal log-likelihood — logLik.dsem","text":"object class logLik attributes val log-likelihood df number parameters Returns object class logLik. attributes \"df\" (degrees freedom) giving number (estimated) fixed effects model, abd \"val\" (value) giving marginal log-likelihood. class allows AIC work expected.","code":""},{"path":"/reference/make_dsem_ram.html","id":null,"dir":"Reference","previous_headings":"","what":"Make a RAM (Reticular Action Model) — make_dsem_ram","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"make_dsem_ram converts SEM arrow notation ram describing SEM parameters","code":""},{"path":"/reference/make_dsem_ram.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"","code":"make_dsem_ram(sem, times, variables, quiet = FALSE, remove_na = TRUE)"},{"path":"/reference/make_dsem_ram.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"sem Specification time-series structural equation model structure including lagged simultaneous effects. See Details section make_dsem_ram description times character vector listing set times order variables character vector listing set variables quiet Boolean indicating whether print messages terminal remove_na Boolean indicating whether remove NA values RAM (default) . remove_NA=FALSE might useful exploration diagnostics advanced users","code":""},{"path":"/reference/make_dsem_ram.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"reticular action module (RAM) describing dependencies","code":""},{"path":"/reference/make_dsem_ram.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"RAM specification using arrow--lag notation line RAM specification make_dsem_ram consists four (unquoted) entries, separated commas: 1. Arrow specification: simple formula, form -> B , equivalently, B <- regression coefficient (.e., single-headed directional arrow); <-> variance <-> B covariance (.e., double-headed bidirectional arrow). , B variable names model. name correspond observed variable, assumed latent variable. Spaces can appear freely arrow specification, can number hyphens arrows, including zero: Thus, e.g., ->B, --> B, >B legitimate equivalent. 2. Lag (using positive values): integer specifying whether linkage simultaneous (lag=0) lagged (e.g., X -> Y, 1, XtoY indicates X time T affects Y time T+1), one-headed arrows can lagged. Using positive values indicate lags matches notational convention used package dynlm. 3. Parameter name: name regression coefficient, variance, covariance specified arrow. Assigning name two arrows results equality constraint. Specifying parameter name NA produces fixed parameter. 4. Value: start value free parameter value fixed parameter. given NA (simply omitted), model provide default starting value. Lines may end comment following #. function extends code copied package `sem` licence GPL (>= 2) permission John Fox. Simultaneous autoregressive process simultaneous lagged effects text specifies linkages multivariate time-series model variables \\(\\mathbf X\\) dimensions \\(T \\times C\\) \\(T\\) times \\(C\\) variables. make_dsem_ram parses text build path matrix \\(\\mathbf{P}\\) dimensions \\(TC \\times TC\\), element \\(\\rho_{k_2,k_1}\\) represents impact \\(x_{t_1,c_1}\\) \\(x_{t_2,c_2}\\), \\(k_1=T c_1+t_1\\) \\(k_2=T c_2+t_2\\). path matrix defines simultaneous equation $$ \\mathrm{vec}(\\mathbf X) = \\mathbf P \\mathrm{vec}(\\mathbf X) + \\mathrm{vec}(\\mathbf \\Delta)$$ \\(\\mathbf \\Delta\\) matrix exogenous errors covariance \\(\\mathbf{V = \\Gamma \\Gamma}^t\\), \\(\\mathbf \\Gamma\\) Cholesky exogenous covariance. simultaneous autoregressive (SAR) process results \\(\\mathbf X\\) covariance: $$ \\mathrm{Cov}(\\mathbf X) = \\mathbf{(- P)}^{-1} \\mathbf{\\Gamma \\Gamma}^t \\mathbf{((- P)}^{-1})^t $$ Usefully, computing inverse-covariance (precision) matrix \\(\\mathbf{Q = V}^{-1}\\) require inverting \\(\\mathbf{(- P)}\\): $$ \\mathbf{Q} = (\\mathbf{\\Gamma}^{-1} \\mathbf{(- P)})^t \\mathbf{\\Gamma}^{-1} \\mathbf{(- P)} $$ Example: univariate first-order autoregressive model simultaneous autoregressive (SAR) process across variables times allows user specify simutanous effects (effects among variables within year \\(T\\)) lagged effects (effects among variables among years \\(T\\)). one example, consider univariate first-order autoregressive process \\(T=4\\). independent errors. specified passing sem = \"X -> X, 1, rho \\n X <-> X, 0, sigma\" make_dsem_ram. parsed RAM: Rows RAM heads=1 interpreted construct path matrix \\(\\mathbf P\\), column \"\" RAM indicates column number matrix, column \"\" RAM indicates row number matrix: $$ \\mathbf P = \\begin{bmatrix} 0 & 0 & 0 & 0 \\\\ \\rho & 0 & 0 & 0 \\\\ 0 & \\rho & 0 & 0 \\\\ 0 & 0 & \\rho & 0\\\\ \\end{bmatrix} $$ rows heads=2 interpreted construct Cholesky exogenous covariance \\(\\mathbf \\Gamma\\) column \"parameter\" RAM associates nonzero element two matrices element vector estimated parameters: $$ \\mathbf \\Gamma = \\begin{bmatrix} \\sigma & 0 & 0 & 0 \\\\ 0 & \\sigma & 0 & 0 \\\\ 0 & 0 & \\sigma & 0 \\\\ 0 & 0 & 0 & \\sigma\\\\ \\end{bmatrix} $$ two estimated parameters \\(\\mathbf \\beta = (\\rho, \\sigma) \\). results covariance: $$ \\mathrm{Cov}(\\mathbf X) = \\sigma^2 \\begin{bmatrix} 1 & \\rho^1 & \\rho^2 & \\rho^3 \\\\ \\rho^1 & 1 + \\rho^2 & \\rho^1 (1 + \\rho^2) & \\rho^2 (1 + \\rho^2) \\\\ \\rho^2 & \\rho^1 (1 + \\rho^2) & 1 + \\rho^2 + \\rho^4 & \\rho^1 (1 + \\rho^2 + \\rho^4) \\\\ \\rho^3 & \\rho^2 (1 + \\rho^2) & \\rho^1 (1 + \\rho^2 + \\rho^4) & 1 + \\rho^2 + \\rho^4 + \\rho^6 \\\\ \\end{bmatrix} $$ converges stationary covariance AR1 process times \\(t>>1\\): $$ \\mathrm{Cov}(\\mathbf X) = \\frac{\\sigma^2}{1+\\rho^2} \\begin{bmatrix} 1 & \\rho^1 & \\rho^2 & \\rho^3 \\\\ \\rho^1 & 1 & \\rho^1 & \\rho^2 \\\\ \\rho^2 & \\rho^1 & 1 & \\rho^1 \\\\ \\rho^3 & \\rho^2 & \\rho^1 & 1\\\\ \\end{bmatrix} $$ except lower pointwise variance initial times, arises \"boundary effect\". Similarly, arrow--lag notation can used specify SAR representing conventional structural equation model (SEM), cross-lagged (.k.. vector autoregressive) models (VAR), dynamic factor analysis (DFA), many time-series models.","code":""},{"path":"/reference/make_dsem_ram.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Make a RAM (Reticular Action Model) — make_dsem_ram","text":"","code":"# Univariate AR1 sem = \" X -> X, 1, rho X <-> X, 0, sigma \" make_dsem_ram( sem=sem, variables=\"X\", times=1:4 ) #> $model #> path lag name start parameter first second direction #> [1,] \"X -> X\" \"1\" \"rho\" NA \"1\" \"X\" \"X\" \"1\" #> [2,] \"X <-> X\" \"0\" \"sigma\" NA \"2\" \"X\" \"X\" \"2\" #> #> $ram #> heads to from parameter start #> 1 1 2 1 1 #> 2 1 3 2 1 #> 3 1 4 3 1 #> 5 2 1 1 2 #> 6 2 2 2 2 #> 7 2 3 3 2 #> 8 2 4 4 2 #> #> $variables #> [1] \"X\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # Univariate AR2 sem = \" X -> X, 1, rho1 X -> X, 2, rho2 X <-> X, 0, sigma \" make_dsem_ram( sem=sem, variables=\"X\", times=1:4 ) #> $model #> path lag name start parameter first second direction #> [1,] \"X -> X\" \"1\" \"rho1\" NA \"1\" \"X\" \"X\" \"1\" #> [2,] \"X -> X\" \"2\" \"rho2\" NA \"2\" \"X\" \"X\" \"1\" #> [3,] \"X <-> X\" \"0\" \"sigma\" NA \"3\" \"X\" \"X\" \"2\" #> #> $ram #> heads to from parameter start #> 1 1 2 1 1 #> 2 1 3 2 1 #> 3 1 4 3 1 #> 5 1 3 1 2 #> 6 1 4 2 2 #> 9 2 1 1 3 #> 10 2 2 2 3 #> 11 2 3 3 3 #> 12 2 4 4 3 #> #> $variables #> [1] \"X\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # Bivariate VAR sem = \" X -> X, 1, XtoX X -> Y, 1, XtoY Y -> X, 1, YtoX Y -> Y, 1, YtoY X <-> X, 0, sdX Y <-> Y, 0, sdY \" make_dsem_ram( sem=sem, variables=c(\"X\",\"Y\"), times=1:4 ) #> $model #> path lag name start parameter first second direction #> [1,] \"X -> X\" \"1\" \"XtoX\" NA \"1\" \"X\" \"X\" \"1\" #> [2,] \"X -> Y\" \"1\" \"XtoY\" NA \"2\" \"X\" \"Y\" \"1\" #> [3,] \"Y -> X\" \"1\" \"YtoX\" NA \"3\" \"Y\" \"X\" \"1\" #> [4,] \"Y -> Y\" \"1\" \"YtoY\" NA \"4\" \"Y\" \"Y\" \"1\" #> [5,] \"X <-> X\" \"0\" \"sdX\" NA \"5\" \"X\" \"X\" \"2\" #> [6,] \"Y <-> Y\" \"0\" \"sdY\" NA \"6\" \"Y\" \"Y\" \"2\" #> #> $ram #> heads to from parameter start #> 1 1 2 1 1 #> 2 1 3 2 1 #> 3 1 4 3 1 #> 5 1 6 1 2 #> 6 1 7 2 2 #> 7 1 8 3 2 #> 9 1 2 5 3 #> 10 1 3 6 3 #> 11 1 4 7 3 #> 13 1 6 5 4 #> 14 1 7 6 4 #> 15 1 8 7 4 #> 17 2 1 1 5 #> 18 2 2 2 5 #> 19 2 3 3 5 #> 20 2 4 4 5 #> 21 2 5 5 6 #> 22 2 6 6 6 #> 23 2 7 7 6 #> 24 2 8 8 6 #> #> $variables #> [1] \"X\" \"Y\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # Dynamic factor analysis with one factor and two manifest variables # (specifies a random-walk for the factor, and miniscule residual SD) sem = \" factor -> X, 0, loadings1 factor -> Y, 0, loadings2 factor -> factor, 1, NA, 1 X <-> X, 0, NA, 0.01 # Fix at negligible value Y <-> Y, 0, NA, 0.01 # Fix at negligible value \" make_dsem_ram( sem=sem, variables=c(\"X\",\"Y\",\"factor\"), times=1:4 ) #> NOTE: adding 1 variances to the model #> $model #> path lag name start parameter first second #> [1,] \"factor -> X\" \"0\" \"loadings1\" NA \"1\" \"factor\" \"X\" #> [2,] \"factor -> Y\" \"0\" \"loadings2\" NA \"2\" \"factor\" \"Y\" #> [3,] \"factor -> factor\" \"1\" NA \"1\" \"0\" \"factor\" \"factor\" #> [4,] \"X <-> X\" \"0\" NA \"0.01\" \"0\" \"X\" \"X\" #> [5,] \"Y <-> Y\" \"0\" NA \"0.01\" \"0\" \"Y\" \"Y\" #> [6,] \"factor <-> factor\" \"0\" \"V[factor]\" NA \"3\" \"factor\" \"factor\" #> direction #> [1,] \"1\" #> [2,] \"1\" #> [3,] \"1\" #> [4,] \"2\" #> [5,] \"2\" #> [6,] \"2\" #> #> $ram #> heads to from parameter start #> 1 1 1 9 1 #> 2 1 2 10 1 #> 3 1 3 11 1 #> 4 1 4 12 1 #> 5 1 5 9 2 #> 6 1 6 10 2 #> 7 1 7 11 2 #> 8 1 8 12 2 #> 9 1 10 9 0 1 #> 10 1 11 10 0 1 #> 11 1 12 11 0 1 #> 13 2 1 1 0 0.01 #> 14 2 2 2 0 0.01 #> 15 2 3 3 0 0.01 #> 16 2 4 4 0 0.01 #> 17 2 5 5 0 0.01 #> 18 2 6 6 0 0.01 #> 19 2 7 7 0 0.01 #> 20 2 8 8 0 0.01 #> 21 2 9 9 3 #> 22 2 10 10 3 #> 23 2 11 11 3 #> 24 2 12 12 3 #> #> $variables #> [1] \"X\" \"Y\" \"factor\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # ARIMA(1,1,0) sem = \" factor -> factor, 1, rho1 # AR1 component X -> X, 1, NA, 1 # Integrated component factor -> X, 0, NA, 1 X <-> X, 0, NA, 0.01 # Fix at negligible value \" make_dsem_ram( sem=sem, variables=c(\"X\",\"factor\"), times=1:4 ) #> NOTE: adding 1 variances to the model #> $model #> path lag name start parameter first second #> [1,] \"factor -> factor\" \"1\" \"rho1\" NA \"1\" \"factor\" \"factor\" #> [2,] \"X -> X\" \"1\" NA \"1\" \"0\" \"X\" \"X\" #> [3,] \"factor -> X\" \"0\" NA \"1\" \"0\" \"factor\" \"X\" #> [4,] \"X <-> X\" \"0\" NA \"0.01\" \"0\" \"X\" \"X\" #> [5,] \"factor <-> factor\" \"0\" \"V[factor]\" NA \"2\" \"factor\" \"factor\" #> direction #> [1,] \"1\" #> [2,] \"1\" #> [3,] \"1\" #> [4,] \"2\" #> [5,] \"2\" #> #> $ram #> heads to from parameter start #> 1 1 6 5 1 #> 2 1 7 6 1 #> 3 1 8 7 1 #> 5 1 2 1 0 1 #> 6 1 3 2 0 1 #> 7 1 4 3 0 1 #> 9 1 1 5 0 1 #> 10 1 2 6 0 1 #> 11 1 3 7 0 1 #> 12 1 4 8 0 1 #> 13 2 1 1 0 0.01 #> 14 2 2 2 0 0.01 #> 15 2 3 3 0 0.01 #> 16 2 4 4 0 0.01 #> 17 2 5 5 2 #> 18 2 6 6 2 #> 19 2 7 7 2 #> 20 2 8 8 2 #> #> $variables #> [1] \"X\" \"factor\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\" # ARIMA(0,0,1) sem = \" factor -> X, 0, NA, 1 factor -> X, 1, rho1 # MA1 component X <-> X, 0, NA, 0.01 # Fix at negligible value \" make_dsem_ram( sem=sem, variables=c(\"X\",\"factor\"), times=1:4 ) #> NOTE: adding 1 variances to the model #> $model #> path lag name start parameter first second #> [1,] \"factor -> X\" \"0\" NA \"1\" \"0\" \"factor\" \"X\" #> [2,] \"factor -> X\" \"1\" \"rho1\" NA \"1\" \"factor\" \"X\" #> [3,] \"X <-> X\" \"0\" NA \"0.01\" \"0\" \"X\" \"X\" #> [4,] \"factor <-> factor\" \"0\" \"V[factor]\" NA \"2\" \"factor\" \"factor\" #> direction #> [1,] \"1\" #> [2,] \"1\" #> [3,] \"2\" #> [4,] \"2\" #> #> $ram #> heads to from parameter start #> 1 1 1 5 0 1 #> 2 1 2 6 0 1 #> 3 1 3 7 0 1 #> 4 1 4 8 0 1 #> 5 1 2 5 1 #> 6 1 3 6 1 #> 7 1 4 7 1 #> 9 2 1 1 0 0.01 #> 10 2 2 2 0 0.01 #> 11 2 3 3 0 0.01 #> 12 2 4 4 0 0.01 #> 13 2 5 5 2 #> 14 2 6 6 2 #> 15 2 7 7 2 #> 16 2 8 8 2 #> #> $variables #> [1] \"X\" \"factor\" #> #> $times #> [1] 1 2 3 4 #> #> attr(,\"class\") #> [1] \"dsem_ram\""},{"path":"/reference/parse_path.html","id":null,"dir":"Reference","previous_headings":"","what":"Parse path — parse_path","title":"Parse path — parse_path","text":"parse_path copied sem::parse.path","code":""},{"path":"/reference/parse_path.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Parse path — parse_path","text":"","code":"parse_path(path)"},{"path":"/reference/parse_path.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Parse path — parse_path","text":"path text parse","code":""},{"path":"/reference/parse_path.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Parse path — parse_path","text":"Tagged-list defining variables direction specified path coefficient","code":""},{"path":"/reference/parse_path.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Parse path — parse_path","text":"Copied package `sem` licence GPL (>= 2) permission John Fox","code":""},{"path":"/reference/plot.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate dsem — plot.dsem","title":"Simulate dsem — plot.dsem","text":"Plot fitted dsem model","code":""},{"path":"/reference/plot.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate dsem — plot.dsem","text":"","code":"# S3 method for dsem plot(x, y, edge_label = c(\"name\", \"value\"), digits = 2, ...)"},{"path":"/reference/plot.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate dsem — plot.dsem","text":"x Output dsem y used edge_label Whether plot parameter names estimated values digits integer indicating number decimal places used ... arguments passed plot.igraph","code":""},{"path":"/reference/plot.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate dsem — plot.dsem","text":"Invisibly returns output graph_from_data_frame passed plot.igraph plotting.","code":""},{"path":"/reference/plot.dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate dsem — plot.dsem","text":"function coerces output graph plots graph.","code":""},{"path":"/reference/predict.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"predictions using dsem — predict.dsem","title":"predictions using dsem — predict.dsem","text":"Predict variables given new (counterfactual) values data, future past times","code":""},{"path":"/reference/predict.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"predictions using dsem — predict.dsem","text":"","code":"# S3 method for dsem predict(object, newdata = NULL, type = c(\"link\", \"response\"), ...)"},{"path":"/reference/predict.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"predictions using dsem — predict.dsem","text":"object Output dsem newdata optionally, data frame look variables predict. omitted, fitted data used create predictions. desiring predictions fitted data, user must append rows NAs future times. Similarly, desiring predictions given counterfactual values time-series data, individual observations can edited keeping observations original fitted values. type type prediction required. default scale linear predictors; alternative \"response\" scale response variable. Thus Poisson-distributed variable default predictions log-intensity type = \"response\" gives predicted intensity. ... used","code":""},{"path":"/reference/predict.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"predictions using dsem — predict.dsem","text":"matrix predicted values dimensions order corresponding argument newdata provided, tsdata . Predictions provided either link response scale, generated re-optimizing random effects condition MLE fixed effects, given new data.","code":""},{"path":"/reference/print.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Print fitted dsem object — print.dsem","title":"Print fitted dsem object — print.dsem","text":"Prints output fitted dsem model","code":""},{"path":"/reference/print.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print fitted dsem object — print.dsem","text":"","code":"# S3 method for dsem print(x, ...)"},{"path":"/reference/print.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print fitted dsem object — print.dsem","text":"x Output dsem ... used","code":""},{"path":"/reference/print.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Print fitted dsem object — print.dsem","text":"return value, called provide clean terminal output calling fitted object terminal.","code":""},{"path":"/reference/residuals.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate residuals — residuals.dsem","title":"Calculate residuals — residuals.dsem","text":"Calculate deviance response residuals dsem","code":""},{"path":"/reference/residuals.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate residuals — residuals.dsem","text":"","code":"# S3 method for dsem residuals(object, type = c(\"deviance\", \"response\"), ...)"},{"path":"/reference/residuals.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate residuals — residuals.dsem","text":"object Output dsem type type residuals compute (option \"deviance\" \"response\" now) ... used","code":""},{"path":"/reference/residuals.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Calculate residuals — residuals.dsem","text":"matrix residuals, order dimensions argument tsdata passed dsem.","code":""},{"path":"/reference/sea_otter.html","id":null,"dir":"Reference","previous_headings":"","what":"Sea otter trophic cascade — sea_otter","title":"Sea otter trophic cascade — sea_otter","text":"Data used demonstrate test trophic cascades options","code":""},{"path":"/reference/sea_otter.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sea otter trophic cascade — sea_otter","text":"","code":"data(sea_otter)"},{"path":"/reference/simulate.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate dsem — simulate.dsem","title":"Simulate dsem — simulate.dsem","text":"Simulate fitted dsem model","code":""},{"path":"/reference/simulate.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate dsem — simulate.dsem","text":"","code":"# S3 method for dsem simulate( object, nsim = 1, seed = NULL, variance = c(\"none\", \"random\", \"both\"), resimulate_gmrf = FALSE, ... )"},{"path":"/reference/simulate.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate dsem — simulate.dsem","text":"object Output dsem nsim number simulated data sets seed random seed variance whether ignore uncertainty fixed random effects, include estimation uncertainty random effects, include estimation uncertainty fixed random effects resimulate_gmrf whether resimulate GMRF based estimated simulated random effects (determined argument variance) ... used","code":""},{"path":"/reference/simulate.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate dsem — simulate.dsem","text":"Simulated data, either obj$simulate obj compiled TMB object, first simulating new GMRF calling obj$simulate.","code":""},{"path":"/reference/simulate.dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate dsem — simulate.dsem","text":"function conducts parametric bootstrap, .e., simulates new data conditional upon estimated values fixed random effects. user can optionally simulate new random effects conditional upon estimated covariance, simulate new fixed random effects conditional upon imprecision. Note simulate effect states x_tj measurement measurements fitted using family=\"fixed\", unless resimulate_gmrf=TRUE. latter case, GMRF resimulated given estimated path coefficients","code":""},{"path":"/reference/summary.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"summarize dsem — summary.dsem","title":"summarize dsem — summary.dsem","text":"summarize parameters fitted dynamic structural equation model","code":""},{"path":"/reference/summary.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"summarize dsem — summary.dsem","text":"","code":"# S3 method for dsem summary(object, ...)"},{"path":"/reference/summary.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"summarize dsem — summary.dsem","text":"object Output dsem ... used","code":""},{"path":"/reference/summary.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"summarize dsem — summary.dsem","text":"Returns data.frame summarizing estimated path coefficients, containing columns: path parsed path coefficient lag lag, e.g. 1 means predictor time t effects response time t+1 name Parameter name start Start value supplied, NA otherwise parameter Parameter number first Variable path treated predictor second Variable path treated response direction Whether path one-headed two-headed Estimate Maximum likelihood estimate Std_Error Estimated standard error Hessian matrix z_value Estimate divided Std_Error p_value P-value associated z_value using two-sided Wald test","code":""},{"path":"/reference/summary.dsem.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"summarize dsem — summary.dsem","text":"DSEM specified using \"arrow lag\" notation, specifies set path coefficients exogenous variance parameters estimated. Function dsem estimates maximum likelihood value coefficients parameters maximizing log-marginal likelihood. Standard errors parameters calculated matrix second derivatives log-marginal likelihood (\"Hessian matrix\"). However, many users want associate individual parameters standard errors path coefficients specified using \"arrow lag\" notation. task complicated models path coefficients variance parameters specified share single value priori, assigned name NA hence assumed fixed value priori (coefficients parameters assigned value standard error). summary function therefore compiles MLE coefficients (including duplicating values path coefficients assigned value) standard error estimates, outputs table associates user-supplied path parameter names. also outputs z-score p-value arising two-sided Wald test (.e. comparing estimate divided standard error standard normal distribution).","code":""},{"path":"/reference/vcov.dsem.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract Variance-Covariance Matrix — vcov.dsem","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"extract covariance fixed effects, fixed random effects.","code":""},{"path":"/reference/vcov.dsem.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"","code":"# S3 method for dsem vcov(object, which = c(\"fixed\", \"random\", \"both\"), ...)"},{"path":"/reference/vcov.dsem.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"object output dsem whether extract covariance among fixed effects, random effects, ... ignored, method compatibility","code":""},{"path":"/reference/vcov.dsem.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract Variance-Covariance Matrix — vcov.dsem","text":"square matrix containing estimated covariances among parameter estimates model. dimensions dependend upon argument , determine whether fixed, random effects, outputted.","code":""},{"path":"/news/index.html","id":"dsem-101","dir":"Changelog","previous_headings":"","what":"dsem 1.0.1","title":"dsem 1.0.1","text":"CRAN release: 2024-01-18 Fix bug simulate.dsem keep changing interface dsem Update CITATION indicate accepted paper","code":""},{"path":"/news/index.html","id":"dsem-100","dir":"Changelog","previous_headings":"","what":"dsem 1.0.0","title":"dsem 1.0.0","text":"CRAN release: 2023-12-08 Initial public release","code":""}]