diff --git a/inst/help/PenalizedMetaAnalysis.md b/inst/help/PenalizedMetaAnalysis.md new file mode 100644 index 00000000..7be94388 --- /dev/null +++ b/inst/help/PenalizedMetaAnalysis.md @@ -0,0 +1,94 @@ +Penalized Meta-Analysis +=== + +Penalized meta-analysis allows the user to estimate Bayesian horseshoe and lasso penalized (multilevel) random-effect meta-regression models. See Van Lissa and colleagues (2023) for more details. + +### Input +--- +#### Assignment Box +- Effect Size: Effect size of the studies. +- Effect Size Standard Error: Standard errors of the effect sizes per study. Must always be positive. A 95% CI for the effect sizes can be provided instead. +- Method: Specify the regularization method for meta-regression. + - Horseshoe: Regularization using the horseshoe prior. + - Lasso: Regularization using the lasso prior. +- Covariates: Continuous predictor variable(s). If ordinal variables are entered it is assumed that their levels are equidistant. Hence, ordinal variables are treated as continuous predictor variables. +- Factors: Categorical predictors variable(s). Ordinal variables here are treated as categorical predictor variables, thus, the ordinal information is ignored. +- Clustering: Variable indicating the presence of higher level cluster. If selected, a three-level meta-regression is estimated. + +### Model +--- + +#### Components and model terms: + - Components: All the independent variables that can be included in the model. + - Model terms: The independent variables in the model. By default, all the main effects of the specified independent variables are included in the model. To include interactions, click multiple variables (e.g., by holding the ctrl/cmd button on your keyboard while clicking) and drag those into the `Model Terms` box. + +#### Include intercept: +- Include the intercept in the regression model. + +#### Scale predictors +- Scale the continuous predictors. + +### Priors +--- + +#### Horshoe +Available only if horseshoe `Method` is selected. +- Df: Degrees of freedom. +- Scale: Scale. + +#### Lasso +Available only if lasso `Method` is selected. +- Df: Degrees of freedom. +- Df (global): Global degrees of freedom. +- Df (slab): Degrees of freedom for the slab. +- Scale (global): Global scale. +- Scale (slab): Scale for the slab. + + +### Inference +--- +#### Estimates table +- Displays table summarizing the posterior distribution of the model terms. + +#### Heterogeneity table +- Displays table summarizing the posterior distribution of the between-study heterogeneity estimate τ (and its square, τ01). + - Displays table summarizing the posterior distribution of the relative heterogeneity I01. + + +### MCMC Diagnostics +--- +#### Model terms +- Model terms whose MCMC chains can be diagnosed. + +#### Plotted term +- Model term which MCMC chains will be diagnosed. + +#### Plot type +- Different types of MCMC diagnostics plots. + - Traceplot: Traceplot of the individual chains. + - Scatterplot: Scatterplot of two model terms. + - Histogram: Histogram of the posterior samples. + - Density: Overlying densities of samples from each chain. + - Autocorrelations: Average autocorrelations across all chains. + + +### Advanced +--- +#### Estimation settings (MCMC) +- Burnin: Number of iterations reserved for burnin. +- Iterations: Number of iterations reserved for sampling. +- Chains: Number of chains. +- Adapt delta: Average target proposal acceptance of each step. Increasing `Adapt delta` results in better-behaved chains, but also longer fitting times. +- Maximum treedepth: The cap for number of trees evaluated during each iteration. Prevents excessively long execution times. + +#### Repeatability +- Set seed: Gives the option to set a seed for your analysis. Setting a seed will exclude random processes influencing an analysis. Note, however, that the seed may not reproduce the same results across operating systems. + + +### References +--- +- Van Lissa, C. J., van Erp, S., & Clapper, E.-B. (2023). Selecting relevant moderators with Bayesian regularized meta-regression. Research Synthesis Methods, 14(2), 301–322. https://doi.org/10.1002/jrsm.1628 + +### R-packages +--- +- pema diff --git a/inst/help/RobustBayesianMetaAnalysis.md b/inst/help/RobustBayesianMetaAnalysis.md index ea68331e..6087ff42 100644 --- a/inst/help/RobustBayesianMetaAnalysis.md +++ b/inst/help/RobustBayesianMetaAnalysis.md @@ -3,7 +3,7 @@ Robust Bayesian Meta-Analysis Robust Bayesian meta-analysis allows the user to specify a wide range of meta-analytic models, combine their estimates using model averaging, and quantify evidence for different hypotheses using Bayes factors. The analysis allows the user to specify various prior distributions for effect sizes and heterogeneity and incorporate models correcting for publication bias with selection models and PET-PEESE. -Please note that we updated the model specification of RoBMA models with the version of JASP 0.15 (the analysis is now built on RoBMA 2.0 package). See our new RoBMA preprint for more details (https://doi.org/10.31234/osf.io/kvsp7). +Please note that we updated the model specification of RoBMA models with the version of JASP 0.15. The default model specification corresponds to RoBMA-PSMA described in Bartoš and colleagues (2023). ### Input --- @@ -27,10 +27,10 @@ The input supplied as standardized effect sizes are internally transformed to Fi The direction of the expected effect size (the publication bias adjusted models with one-sided weight functions and PET-PEESE publication bias adjustments are not symmetrical around zero). #### Model type -Either one of the three pre-specified model types corresponding to the models introduced in Bartoš et al. (2021) and Maier, Bartoš & Wagenmakers (in press), or a custom ensemble. -- RoBMA-PSMA corresponds to the 36 model ensemble that combines selection models and PET-PEESE adjustment for publication bias adjustment component (from Bartoš et al., 2021) -- RoBMA-PP corresponds to the 12 model ensemble that uses PET-PEESE adjustment for publication bias adjustment component (from Bartoš et al., 2021) -- RoBMA-original corresponds to the 12 model ensemble that uses two two-sided weight functions for publication bias adjustment component (from Maier, Bartoš & Wagenmakers, in press) +Either one of the three pre-specified model types corresponding to the models introduced in Bartoš et al. (2023) and Maier, Bartoš & Wagenmakers (2023), or a custom ensemble. +- RoBMA-PSMA corresponds to the 36 model ensemble that combines selection models and PET-PEESE adjustment for publication bias adjustment component (from Bartoš et al., 2023) +- RoBMA-PP corresponds to the 12 model ensemble that uses PET-PEESE adjustment for publication bias adjustment component (from Bartoš et al., 2023) +- RoBMA-original corresponds to the 12 model ensemble that uses two two-sided weight functions for publication bias adjustment component (from Maier, Bartoš & Wagenmakers, 2023) - Custom allows specifying a custom model ensemble under the `Models` section #### Prior scale @@ -238,9 +238,9 @@ Balances the prior model probability across models with the same combinations of ### References --- -- Maier, M., Bartoš, F., & Wagenmakers, E. J. (in press). Robust Bayesian meta-analysis: Addressing publication bias with model-averaging. Psychological Methods. https://doi.org/10.31234/osf.io/u4cns -- Bartoš, F., Maier, M., Wagenmakers, E. J., Doucouliagos, H., & Stanley, T. D. (2021). No need to choose: Robust Bayesian meta-analysis with competing publication bias adjustment methods. https://doi.org/10.31234/osf.io/kvsp7 -- Bartoš, F., Maier, M., Quintana, D. S., & Wagenmakers, E. J. (2020). Adjusting for publication bias in JASP & R — selection models, PET-PEESE, and robust Bayesian meta-analysis. https://doi.org/10.31234/osf.io/75bqn +- Maier, M., Bartoš, F., & Wagenmakers, E. J. (2023). Robust Bayesian meta-analysis: Addressing publication bias with model-averaging. Psychological Methods 28 (1), 107-122. https://doi.org/10.1037/met0000405 +- Bartoš, F., Maier, M., Wagenmakers, E. J., Doucouliagos, H., & Stanley, T. D. (2023). Robust Bayesian meta-analysis: Model-averaging across complementary publication bias adjustment methods. Research Synthesis Methods 14 (1), 99-116. https://doi.org/10.1002/jrsm.1594 +- Bartoš, F., Maier, M., Quintana, D. S., & Wagenmakers, E. J. (2022). Adjusting for publication bias in JASP & R: Selection models, PET-PEESE, and robust Bayesian meta-analysis. Advances in Methods and Practices in Psychological Science 5 (3), 1-19. https://doi.org/10.1177/25152459221109259 ### R-packages ---