[R-sig-ME] p values for factor effect in lmer

Jonathan Baron baron at psych.upenn.edu
Thu Dec 13 18:45:37 CET 2007

Suppose I have something like

l1 <- lmer(y ~ x + f + (1 | s))

where f is a factor with levels 1, 2, 3.

l1 itself lists the individual values, f1, f2, f3, not the overall
effect of f.

anova(l1) gives the overall effect of f, but no p values.

pvals.fnc(l1) from the languageR package is like l1 itself.  But I
want to know whether the factor has a significant effect.

I could do

l2 <- lmer(y ~ x + (1 | s))

but, in my experience so far, this method gives answers that don't
look right and don't agree with other methods.  I have not been
trusting it to provide p values.  pvals.fnc(), by contrast, usually
gives very sensible values that agree with other approaches.

Can anyone suggest another approach?  ("Wait until the next version of
..." is a legitimate answer, if it is true.)

Or maybe I should trust anova(l1,l2) after all.

Jonathan Baron, Professor of Psychology, University of Pennsylvania
Home page: http://www.sas.upenn.edu/~baron

More information about the R-sig-mixed-models mailing list