[R-sig-ME] p-values vs likelihood ratios

[Mike Lawrence]

On Mon, Feb 21, 2011 at 9:39 AM, Jarrod Hadfield <j.hadfield at ed.ac.uk> wrote:
> Also, comparing REML likelihoods between models where the fixed effects have
> changed is, I think, inappropriate because the data have essentially
> changed. You should be OK if you specify REML=FALSE.

Ah yes, an important point of which I was aware (and implement in
ez::ezMixed), but neglected to make explicit in the examples I
provided.