fix varianceBySigma bracket error

.Rbuildignore

@@ -1,3 +1,4 @@
+^CRAN-RELEASE$
 ^.*\.Rproj$
 ^\.Rproj\.user$      
 ^README\.Rmd$

.gitignore

@@ -27,3 +27,4 @@ vignettes/*.pdf
 .Rproj.user
 *.Rproj
 README.html
+CRAN-RELEASE

DESCRIPTION

@@ -15,5 +15,5 @@ Suggests: testthat, knitr,  dplyr, ggplot2, mvtnorm, purrr, tidyr
 VignetteBuilder: knitr
 License: GPL-2
 LazyData: true
-RoxygenNote: 6.0.1
+RoxygenNote: 6.1.1
 URL: https://github.com/bgctw/lognorm

NEWS.md

@@ -3,7 +3,7 @@
 - Sum of lognormals: if all values are gap-filled, assume multiplicative
    standard deviation (sigmaStar) of the sum to be the mean of given sigmaStar.
    Before took it assumed to be the maximum. But this led to larger confidence 
-   boundsas compared to using the normal assumption.
+   bounds compared to using the normal assumption.
 
 - Estimate parameters from various statistics, such as mean and an
     upper quantile value, i.e. a practical maximum.
@@ -15,12 +15,12 @@ function `seCor`.
 
 # lognorm 0.1.3
 
-EstimateSumLognormal by default now returns NA if there are NA values in terms.
-New argument na.rm = TRUE allows for previous behaviour of neglecting those terms.
+estimateSumLognormal by default now returns NA if there are NA values in terms.
+New argument na.rm = TRUE allows for previous behavior of neglecting those terms.
 
 # lognorm 0.1.2
 
-- Computate of effective number of observations
+- Compute effective number of observations
   - fixed bracketing bug in formula
   - better dealing with NA at begin or end
 

R/autocorr.R

@@ -233,7 +233,7 @@ seCor <- function(
   ##details<< Computation follows 
   ## https://stats.stackexchange.com/questions/274635/calculating-error-of-mean-of-time-series.
   ## 
-  ##details<< The default uses empiricial autocorrelation
+  ##details<< The default uses empirical autocorrelation
   ## estimates from the supplied data up to first negative component.
   ## For short series of \code{x} it is strongly recommended to to
   ## provide \code{effCov} that was estimated on a longer time series.

cran-comments.md

@@ -1,17 +1,17 @@
 ## Note
-Added support for the common use case of computing standard error of an 
-autocorrelated time series.
+Added support for computing moments and statistics.
 
 ## Test environments
 * local R 3.4.4 on Mint17 64bit
 * Ubuntu 14.04.5 LTS (on travis-ci), 14.04
 * win-builder (devel)
+* R-Hub
 
 ## R CMD check results
 There were no ERRORs nor WARNINGs 
 
-One note on Possibly mis-spelled words in DESCRIPTION
-  caused by correct "et al." of a citation.
+Note on wrong doi: I checked that the doi correctly resolves
+http://dx.doi.org/10.1641%2F0006-3568(2001)051[0341%3Alndats]2.0.co%3B2
 
 
 

inst/docu/varianceBySigmaAndExpected.Rmd

@@ -0,0 +1,60 @@
+---
+title: "testing variance computation with expected value and sigma"
+output: html_notebook
+---
+
+Formula for Variance given sigma at log scale and expected value
+
+We start with formulas for variance and mean.
+$$
+V =  (e^{\sigma^2} - 1) e^{2\mu + \sigma^2} \\
+m = e^{\mu + \sigma^2/2}
+$$
+mean formula resolved for $\mu$ and substituted into V.
+$$
+\mu = log_m - \sigma^2/2 \\
+V =  (e^{\sigma^2} - 1) e^{2(log_m - \sigma^2/2) + \sigma^2} \\
+V =  (e^{\sigma^2} - 1) e^{2log_m - \sigma^2 + \sigma^2} \\
+V =  (e^{\sigma^2} - 1) e^{2log_m} \\
+V =  (e^{\sigma^2} - 1) (e^{log_m})^2 \\
+V =  (e^{\sigma^2} - 1) m^2 \\
+$$
+
+
+```{r}
+n = 1e4
+sigma = log(1.2)
+#sigma = log(1.41)
+mu = log(10)
+logR = rnorm(n, mu, sigma)
+R = exp(logR)
+meanR = mean(R)
+sdR = sd(R)
+V2 = (exp(sigma^2) - 1)*exp(2*mu + sigma^2)
+#
+m = exp(mu + sigma^2/2)
+V = (exp(sigma^2) - 1)*m^2
+c(meanR, m)
+c(sdR, sqrt(V), sqrt(V2))
+```
+```{r}
+xPred <- seq(-2,22,length.out = 101) 
+plot(density(R), xlim = c(-2,22), lty = "dotted")
+abline(v = meanR)
+lines(dnorm(xPred, m, sqrt(V))~xPred, col = "blue", lty = "dashed")
+lines(dlnorm(xPred, mu, sigma)~xPred, col = "green")
+```
+```{r}
+df <- data.frame(nu = c(0.05,0.1,0.2,0.5,1,2,5,10,20))
+df$sigmaStar = sqrt(df$nu^2 + 1)
+df
+```
+```{r}
+plot(sigmaStar ~ nu, df[1:3,])
+```
+
+```{r}
+sigmaStar <- 1.2
+(nu <- sqrt(sigmaStar^2 - 1))
+```
+

man/seCor.Rd

@@ -14,7 +14,7 @@ first entry is the variance. See \code{\link{computeEffectiveAutoCorr}}}
 https://stats.stackexchange.com/questions/274635/calculating-error-of-mean-of-time-series.
 
 
-The default uses empiricial autocorrelation
+The default uses empirical autocorrelation
 estimates from the supplied data up to first negative component.
 For short series of \code{x} it is strongly recommended to to
 provide \code{effCov} that was estimated on a longer time series.}