lmerTest (version 2.0-6)

lmerTest-package: The package performs different kinds of tests on lmer objects, such as F tests of type 3/type 1 hypotheses for the fixed part, likelihood ratio tests for the random part, least squares means (population means) and differences of least squares means for the factors of the fixed part with corresponding plots. The package also provides with a function step, that preforms backward elimination of non-significant effects, starting from the random effects, and then fixed ones.

Description

The package provides anova function, that gives data frame similar to what gives lme4 package but with p-values calculated from F statistics of type 3/type 1 hypotheses. There are two options for denominator degrees of freedom of F statistics: "Satterthwaite" and "Kenward-Roger". The calculation of anova with Kenward-Roger's approximation is based on function from pbkrtest package, the calculation of Satterthwaite's approximation is based on SAS proc mixed theory (see reference).The type 3 hypothesis (marginal) is calculated according to SAS theory (SAS Institute Inc., 1978). The package also provides summary function, which gives the same as lme4 package summary function but with p-values and degrees of freedom added for the t-test (based on Satterthwaite approximation for denominator degrees of freedom). The tests on random effects are performed using likelihood ratio tests.

Details

ll{ Package: lmerTest Type: Package Version: 1.0 Date: 2012-01-10 License: GPL } The calculation of statistics for the fixed part was developed according to SAS Proc Mixed Theory (see reference).

References

SAS Technical Report R-101 1978 Tests of Hypotheses in Fixed-Effects Linear Models Copyright (C) (SAS Institute Inc., Cary, NC, USA) Goodnight, J.H. 1976 General Linear Models Procedure (S.A.S. Institute, Inc.) Schaalje G.B., McBride J.B., Fellingham G.W. 2002 Adequacy of approximations to distributions of test Statistics in complex mixed linear models

Examples

#import lmerTest package
library(lmerTest)

# an object of class merModLmerTest
m <- lmer(Informed.liking ~ Gender+Information+Product +(1|Consumer), data=ham)

# gives summary of lmer object. The same as of class merMod but with 
# additional p-values calculated based on Satterthwate's approximations
summary(m)

# anova table the same as of class merMod but with additional F statistics and 
#  and denominator degrees of freedom and 
# p-values calculated based on Satterthwaite's approximations
anova(m)

# anova table the same as of class merMod but with additional F statistics and  
# denominator degrees of freedom and 
# p-values calculated based on Kenward-Roger's approximations
anova(m, ddf="Kenward-Roger")

# anova table of class merMod 
anova(m, ddf="lme4")

# backward elimination of non-significant effects of model m
st <- step(m)

plot(st)