Bug fix for three-level models

This release fixes a bug in the Fisher_info() function that returned the information matrix in an incorrect order for models with three or more sets of random effects or models with multiple sets of random effects at the same level. Additional changes:

• Removed cnvg_warn argument from the g_mlm() function.
• Updated the unit tests related to Lambert et al. (2006) data.

Bug fix

This release includes a fix for a bug in extract_varcomp() that caused some variance components to be dropped if the variables involved in the random effects formula involved special characters such as ., (, ), or ^.

Separate level-1 variance components

• Added an option to return separate level-1 variance components for models that use weights = varIdent(form = ~ 1 | Stratum).
• Generalized the g_mlm() function to allow use of separate models for the numerator and denominator of the effect size.
• Modified the stored results of g_mlm() so that the returnModel argument is no longer necessary.
• Fixed a bug in handling of models with missing observations that have na.action of na.exclude().
• Fixed a bug in internal functions for constructing level-1 variance covariance structures in models with multi-variate structure (i.e., models with non-null modelStruct$corStruct).

bug fixes and updates

• Corrected a bug leading to errors in Fisher_info() with models that have more than two levels.
• Added informative errors for summary.g_mlm() if setting returnModel = FALSE.
• Updated the formula for standard error of effect size in the vignette to match the method in g_mlm().
• Added tests for handling models with missing outcome and/or covariate values.

Updates

• Fixed the "additional issues" in the unit tests identified by the CRAN package checks.
• Added pkgdown.

First CRAN release

First release on CRAN.