re: st: covariance structure name?

[David Airey <david.airey@vanderbilt.edu>]

> Please note that this is a stats question and not a Stata question;
> which is the purpose of the list. If you have a stats question try
> MedStats.
>
> It would seem you have a autoregressive structure. The autoregressive
> correlation model formalizes the idea that the magnitude of
> correlation among observations "decays" or "drop off" as they become
> farther apart.
>
> Raphael
No, I did not frame the question with regard to a specific Stata  
command...but your answer did help in that regard.

The six measures were from consecutive minutes over a six minute  
interval. Mouse behavior was measured. Multiple mice are each tested  
under a saline and two drugs. Replicate mice are nested in 20  
isogenic lines, where lines are nested in two genotype classes. The  
effect of genotype on drug response was of interest. I used a sum of  
the behavior measure across all six minutes, and ran the following  
model:

xi: xtmixed sum_6_min i.drug*i.genotype || strain: || mouse:

This is fine since I was not really interested in the time aspect of  
the 6 measures, but I was thinking about running the model on the raw  
data. My question was really about the setting of the cov() option in  
xtmixed should I decide to try this.

I thought autoregressive had the same variances along the diagonal.  
In any case, I think it is moot since I only see unstructured as an  
alternative in xtmixed.

Where is MedStats? I could not find a google link that seemed to make  
sense. I would like a good statistics listserv for questions about  
statistics generally.

Thanks.

-Dave
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/