[R-sig-ME] QQ Norm

[Douglas Bates]

On Dec 7, 2007 12:15 PM, Douglas Bates <bates at stat.wisc.edu> wrote:
> On Dec 7, 2007 11:18 AM, Roberts, Kyle <kyler at mail.smu.edu> wrote:
> > Just a simple yes or no.  Is there any difference between:
> >
> > qqnorm(resid(lmer.out))
> > and
> > qqmath(~resid(lmer.out))
>
> That's a bad question for which to say that you want a simple yes/no
> because the answer is obciously yes and now you are going to need to
> ask what the differences are.

I was being facetious with that answer.  There are obvious differences
in that qqnorm is a standard graphics function and qqmath is
lattice/grid.  Other than that there isn't really a difference.

The only class of objects for which the lme4 package defines a special
qqmath method is the ranef classes, such as ranef.mer and ranef.lmer.
For those classes the method creates a form of "caterpillar plot" if
you ask for the posterior variances (which, if I had it to do over
again, I would call "conditional variances") to be returned (i.e.
postVar = TRUE).

By the way, it is now possible to omit the ~ in the first argument for
qqmath so you could write

qqmath(resid(lmer.out))

Whether it is advisable to do so is something I still haven't decided
for myself.