Message-ID: <6486FDE3-2E6C-4FF2-84A6-AF013CBEC5C0@kagi.com>
Date: 2008-01-14T21:26:53Z
From: Ken Beath
Subject: Some Basic lmer Questions
In-Reply-To: <20080114205024.GJ995@ms.unimelb.edu.au>
On 15/01/2008, at 7:50 AM, Andrew Robinson wrote:
> Hi Kevin,
>
> welcome to the wonderful world of mixed-models! I heartily recommend
> that you read the following books: Gelman and Hill, and Pinehiro and
> Bates.
>
For linear mixed models I would also suggest Verbeke and Molenberghs
"Linear Mixed Models for Longitudinal Data"
> On Mon, Jan 14, 2008 at 03:21:20PM -0500, Kevin E. Thorpe wrote:
>>
>> Finally, (and feel free to point me at suitable references)
>> how does one determine the appropriate correlation structure
>> to use in these models?
While you can make life more complicated AIC is a reasonable method.
This is criticised as it is not the same as statistical significance
but all we need is a reasonable model for the correlation so that
inferences about the other parameters are correct.
Ken