Meta regression (#21)

* down-weighting

* rename main.R->RoBMA.R

* RoBMA.reg

* summary functions

* fixes

* rename main.R->RoBMA.R

* RoBMA.reg

* summary functions

* fixes

* additional fixes

* updating unit tests

* plot function

* updating diagnostics to BayesTools

* fix plots

* do_not_fit feature

* do_not_fit feature (2)

* fixing the weights parameterization

* add the students data set

* add the students data set (1)

* fixes

* vignette updates (1)

* additional models

* additional models II

* additional models (3)

* updates

* updating the vignette --still not done yet

* almost finished vignette draft

* more model results

* Create fit_wPSMA_reg10.RDS

* updating vignettte

* Update .gitignore

* Create fit_PSMA_reg10.RDS

* Create fit_3PP_reg10

* changing the multivariate normal C++ implementation to the same one as in bugs

* fit_3PP_reg10 (non-converged)

* fix multivariate

* updating hierarchical model specification & adding a vignette

* Update fit_3PP_reg10.RDS

* fixes and update unit tests

* add AMPPS vignette

* updates

* updates

* include Lui2-15 data data set & update the vignette

* adding inverse_sqrt weights

* fixing data input checks

* ggplot2 unit test updates

* Update utilities.R

* update with BayesTools

* remove old meta-regression viggnete & change defaults priors for factors/covariates

* add predictor scaling warning

* Update plots.R

* Update DESCRIPTION

* Update RoBMA.R

* check_setup.reg() + fixes

* unit tests 3-5

* unit test 6

* unit test 7

* coverage on windows (?)

* Revert "coverage on windows (?)"

This reverts commit 78b4880.

* reparameterization of the `weighted` models experimental feature (vol I)

* undo reparameterization of the `weighted` models experimental feature

* update documentation

* Update check-input-and-settings.R

* improving regression specification

* update documentation

* Update test-2-distributions.R

* add unit tests for regression parameters & update R4.3 tests

* updating summary.RoBMA.Rd

* update with BayesTools

* Update test-5-methods.R

* marginal functions + unit test update

* unit tests + small fixes

* change coverage tests

* Update utilities.R

* add regression CRAN test and mute mesages

* small updates

* update README

* Update prior_informed.Rd

* update actions

Author: FBartos

.RData

@@ -22,7 +22,7 @@ jobs:
       fail-fast: false
       matrix:
         config:
-          - {os: ubuntu-20.04,   r: 'release', rspm: "https://packagemanager.rstudio.com/cran/__linux__/focal/latest"}
+          - {os: windows-latest, r: 'release'}
 
     env:
       GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/test-coverage.yaml

@@ -17,4 +17,7 @@ RoBMA.Rcheck
 .Rprofile
 /doc/
 /Meta/
+models/MetaRegression/fit_wPSMA_reg10/*.RDS
+models/MetaRegression/fit_PSMA_reg10/*.RDS
+models/MetaRegression/fit_3PP_reg10/*.RDS
 models/MetaRegression/*

.gitignore

@@ -1,6 +1,6 @@
 Package: RoBMA
 Title: Robust Bayesian Meta-Analyses
-Version: 2.3.2
+Version: 3.0.0
 Maintainer: František Bartoš <f.bartos96@gmail.com>
 Authors@R: c( 
     person("František", "Bartoš",     role = c("aut", "cre"),
@@ -21,9 +21,9 @@ Description: A framework for estimating ensembles of meta-analytic models
     combine the competing meta-analytic models into a model ensemble, weights 
     the posterior parameter distributions based on posterior model probabilities 
     and uses Bayes factors to test for the presence or absence of the
-    individual components (e.g., effect vs. no effect; Bartoš et al., 2021, 
-    <doi:10.31234/osf.io/kvsp7>; Maier, Bartoš & Wagenmakers, in press, 
-    <doi:10.31234/osf.io/u4cns>). Users can define a wide range of non-informative 
+    individual components (e.g., effect vs. no effect; Bartoš et al., 2022, 
+    <doi:10.1002/jrsm.1594>; Maier, Bartoš & Wagenmakers, 2022, 
+    <doi:10.1037/met0000405>). Users can define a wide range of non-informative 
     or informative prior distributions for the effect size, heterogeneity, 
     and publication bias components (including selection models and PET-PEESE). 
     The package provides convenient functions for summary, visualizations, and 
@@ -40,10 +40,9 @@ NeedsCompilation: yes
 Depends:
     R (>= 4.0.0)
 Imports:
-    BayesTools (>= 0.2.0),
+    BayesTools (>= 0.2.14),
     runjags,
     rjags,
-    coda,
     stats,
     graphics,
     mvtnorm,
@@ -53,8 +52,11 @@ Imports:
     ggplot2
 Suggests:
     parallel,
-    rstan,
     metaBMA,
+    metafor,
+    weightr,
+    fixest,
+    metadat,
     testthat,
     vdiffr,
     knitr,

DESCRIPTION

@@ -2,6 +2,7 @@
 
 S3method(plot,RoBMA)
 S3method(print,RoBMA)
+S3method(print,marginal_summary.RoBMA)
 S3method(print,summary.RoBMA)
 S3method(summary,RoBMA)
 S3method(update,RoBMA)
@@ -12,29 +13,40 @@ export(OR2z)
 export(RoBMA)
 export(RoBMA.get_option)
 export(RoBMA.options)
+export(RoBMA.reg)
 export(check_RoBMA)
 export(check_setup)
+export(check_setup.reg)
 export(combine_data)
+export(contr.meandif)
+export(contr.orthonormal)
 export(d2OR)
 export(d2logOR)
 export(d2r)
 export(d2z)
 export(diagnostics)
+export(diagnostics_autocorrelation)
+export(diagnostics_density)
+export(diagnostics_trace)
 export(dwnorm)
 export(forest)
 export(interpret)
 export(is.RoBMA)
+export(is.RoBMA.reg)
 export(logOR2OR)
 export(logOR2d)
 export(logOR2r)
 export(logOR2z)
+export(marginal_plot)
+export(marginal_summary)
 export(n_d)
 export(n_r)
 export(n_z)
 export(plot_models)
 export(prior)
 export(prior_PEESE)
 export(prior_PET)
+export(prior_factor)
 export(prior_informed)
 export(prior_none)
 export(prior_weightfunction)
@@ -69,6 +81,7 @@ export(z2r)
 importFrom(BayesTools,is.prior)
 importFrom(BayesTools,is.prior.PEESE)
 importFrom(BayesTools,is.prior.PET)
+importFrom(BayesTools,is.prior.factor)
 importFrom(BayesTools,is.prior.none)
 importFrom(BayesTools,is.prior.point)
 importFrom(BayesTools,is.prior.simple)

NAMESPACE

@@ -1,3 +1,17 @@
+## version 3.0
+### Features
+- meta-regression with `RoBMA.reg()` function
+- posterior marginal summary and plots for the `RoBMA.reg` models with `summary_marginal()` and `plot_marginal()` functions
+- new vignette on hierarchical Bayesian model-averaged meta-analysis
+- new vignette on robust Bayesian model-averaged meta-regression
+- adding vignette from AMPPS tutorial
+- faster implementation of JAGS multivariate normal distribution (based on the BUGS JAGS module)
+- incorporating `weight` argument in the `RoBMA` and `combine_data` functions in order to pass `custom` likelihood weights
+- ability to use inverse square weights in the weighted meta-analysis by setting a `weighted_type = "inverse_sqrt"` argument 
+
+### Changes
+- reworked interface for the hierarchical models. Prior distributions are now specified via the `priors_hierarchical` and `priors_hierarchical_null` arguments instead of `priors_rho` and `priors_rho_null`. The model summary now shows `Hierarchical` component summary.
+
 ## version 2.3.2
 ### Fixes
 - suppressing start-up message 

NEWS.md

@@ -24,7 +24,7 @@
 ##' further questions.
 ##'
 ##' @references \insertAllCited{}
-##' @importFrom BayesTools is.prior is.prior.none is.prior.point is.prior.simple is.prior.PET is.prior.PEESE is.prior.weightfunction
+##' @importFrom BayesTools is.prior is.prior.none is.prior.point is.prior.simple is.prior.factor is.prior.PET is.prior.PEESE is.prior.weightfunction
 ##' @importFrom Rdpack reprompt
 ##' @importFrom rlang .data
 "_PACKAGE"