Message-ID: <4A4218F7.4070108@ufl.edu>
Date: 2009-06-24T12:15:51Z
From: Ben Bolker
Subject: lmer vs glmmPQL
In-Reply-To: <FD219D38-EC84-4154-90D5-D4654FE2E41F@imperial.ac.uk>
Federico Calboli wrote:
> On 23 Jun 2009, at 22:46, Ken Beath wrote:
>> This seems to results from the use of a t-test with few df in glmmPQL
>> and z in lmer. z seems fine to me. What is more of a problem is that
>> your random effects variance is effectively 0. There are only 3 blocks
>> so fitting a random effects model will be difficult and appears
>> unnecessary.
>
>
> That was a sample dataset so I could see what kind of data I had to
> deal with, the 'real' hing should have a variance > 0 for the random
> effect. My philosphycal issue was, given such a relatively
> straightforward model, should I be more (glmmPQL) or less (lmer)
> conservative?
>
> Best,
>
> Federico
My take would be to pick lmer over glmmPQL every time, provided
it can handle your problem -- in general it should be more accurate.
Picking on the basis of more or less conservative for any given problem
feels biased.
Ben Bolker