Skip to content

Help: Specifying 'correct' degrees of freedom for within-subject factor in nlme repeated measures ANOVA

2 messages · Stephanie Avery-Gomm, Mollie Brooks

#
Specifying denominator degrees of freedom in models with random effects is still controversial. I think the bottom line is that if you want to get it right, do MCMC or bootstrapping instead.

This FAQ page usually keeps up on the latest best practices in light of the difficulties http://glmm.wikidot.com/faq
Scroll down to the section on Degrees of Freedom. It talks about df in the context of lme4, but the same ideas should apply to lme.

You would probably get more responses on the R-sig-mixed list.
cheers,
Mollie
On 26 Jul 2012, at 12:55 PM, Stephanie Avery-Gomm wrote: