glmmVCtest: glmmVCtest: Testing variance components in generalized linear...

In stchen3/glmmVCtest: Testing variance components in generalized linear mixed models (GLMMs)

Description

The glmmVCtest package implements four methods for testing a single random effect (variance components) in a generalized linear mixed model (GLMM):

• Normalized restricted likelihood ratio test (nRLRT): Chen et al. (2019) (recommended).

• Asymptotic-normalized restricted likelihood ratio test (as-nRLRT).

• Asymptotic likelihood ratio test (asLRT): Molenberghs and Verbeke (2007).

• Normalized Score test (nScore): (Zhang and Lin 2003).

We recommend using the nRLRT method. This method uses penalized quasi-likelihood (PQL) to approximate the GLMM with a working linear mixed model (LMM), then compares the RLRT statistic to its finite-sample null distribution for hypothesis testing. Compared to the other methods, it a) applies to a greater range of models, b) has better type I error rates, and c) has higher power.

Details

These methods are intended for testing a single variance component in a GLMM for responses from an exponential family distribution. This is limited to the distributions that can be estimated using MASS::glmmPQL or lme4::glmer. Currently, normal responses are not supported; testing can be done directly with the RLRsim package.

glmmVCtest functions

• asymptotic.null: asymptotic null distribution from Self and Liang (1987)

• nScore: normalized Score test (Zhang and Lin 2003)

• glmmPQL.mod: estimate a GLMM using PQL (modification of MASS::glmmPQL)

• test.asnRLRT: approximate-normalized RLRT

• test.nRLRT: normalized RLRT (recommended)

• test.asLRT: asymptotic LRT (Molenberghs and Verbeke 2007)

See Also

Chen, S. T., Xiao, L., Staicu, A. M. (in prep). Restricted Likelihood Ratio Tests for Variance Components in Generalized Linear Models.

Molenberghs, G. and Verbeke, G. (2007). Likehood ratio, score, and wald tests in a constrained parameter space. The American Statistician 61, 22–27.

Zhang, D. and Lin, X. (2003). Hypothesis testing in semiparametric additive mixed models. Biostatistics 4, 57–74.

stchen3/glmmVCtest documentation built on May 23, 2019, 2:48 p.m.