Confidence Intervals for Robust and Classical Linear Mixed Model Estimators

The main function calculates confidence intervals (CI) for Mixed Models, utilizing both classical estimators from the lmer() function in the 'lme4' package and robust estimators from the rlmer() function in the 'robustlmm' package, as well as the varComprob() function in the 'robustvarComp' package. Three methods are available: the classical Wald method, the wild bootstrap, and the parametric bootstrap. Bootstrap methods offer flexibility in obtaining lower and upper bounds through percentile or BCa methods. More details are given in Mason, F., Cantoni, E., & Ghisletta, P. (2021) and Mason, F., Cantoni, E., & Ghisletta, P. (2024) .

confintROB

Contains a set of R functions to compute percentile or BCa confidence intervals based on bootstrap schemes, for linear mixed models estimated with both classical and robust methods.

The scripts allow for balanced and unbalanced data, using rlmer (Koller, 2016) for robust estimation or lmer (Bates et al., 2015) for classical estimation.

Two bootstrap schemes are implemented: the wild bootstrap and the parametric bootstrap.

Both are adapted for object of class lmerMod (with lmer), rlmerMod (with rlmer) or varComprob (varComprob) (Agostinelli & Yohai, 2016).

A comparison of performance with balanced datasets is available in Mason, Cantoni and Ghisletta (2024).

Contents

The main function is confintROB; it is implemented as an extension of the popular confint (confint.merMod) function of the lme4 package.