[R] F tests for random effect models

[Doran, Harold]

I think what you're looking for is in anova() 

> fm1 <- lmer(dv ~ IV ...)
> anova(fm1)



-----Original Message-----
From: r-help-bounces at stat.math.ethz.ch
[mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Jacques VESLOT
Sent: Thursday, October 27, 2005 2:22 AM
To: R-help at stat.math.ethz.ch
Subject: [R] F tests for random effect models

Dear R-users,

My question is how to get right F tests for random effects in random
effect models (I hope this question has not been answered too many times
yet - I didn't find an answer in rhelp archives).

My data are in mca2 (enc.) :

names(mca2)
[1] "Lignee"        "Pollinisateur" "Rendement"

dim(mca2)
[1] 100   3

replications(Rendement ~ Lignee * Pollinisateur, data = mca2)
               Lignee        Pollinisateur Lignee:Pollinisateur
                   20                   10                    2

Of course, summary(aov(Rendement ~ Pollinisateur * Lignee, data = mca2))
gives wrong tests of random effects. But, summary(aov1 <- aov(Rendement
~ Error(Pollinisateur * Lignee), data = mca2)) gives no test at all, and
I have to do it like this :

tab1 <- matrix(unlist(summary(aov1)), nc=5, byrow=T)[,1:3]

Femp <- c(tab1[1:3, 3]/tab1[c(3,3,4), 3])

names(Femp) <- c("Pollinisateur", "Lignee", "Interaction")

1 - pf(Femp, tab1[1:3,1], tab1[c(3,3,4),1])

With "lme4" package (I did'nt succeed in writing a working formula with
lme from "nlme" package), I can "see" standard deviations of random
effects (but don't know how to find them) with :

library(lme4)
summary(lmer(Rendement ~ (1 |Pollinisateur) + (1 | Lignee) + (1 |
Pollinisateur:Lignee), data=mca2))

but I can't get F tests.

Thanks in advance.

Best regards,

Jacques VESLOT