buildmer - Stepwise Elimination and Term Reordering for Mixed-Effects
Regression
Finds the largest possible regression model that will
still converge for various types of regression analyses
(including mixed models and generalized additive models) and
then optionally performs stepwise elimination similar to the
forward and backward effect-selection methods in SAS, based on
the change in log-likelihood or its significance, Akaike's
Information Criterion, the Bayesian Information Criterion, the
explained deviance, or the F-test of the change in R².
