Re: st: Re: metareg w/ clustering?

[Roger Harbord <roger.harbord@bristol.ac.uk>]

-metareg- does meta-analysis regression, aka. meta-regression. It assumes 
you have summary data for an outcome and its variance within each 
group/study, similar to -vwls-, except it estimates an additional 
between-group/study variance component tau-squared. Using regular 
regression with analytic weights would estimate a multiplicative 
overdispersion parameter that multiplies each within-study variance rather 
than an additive component of variance.

-metareg- does not have -cluster- or -robust- options. It would not be 
straightforward to add these to the code as it does not use conventional 
maximum likelihood estimation. I guess it might be possible to use 
-_robust- afterwards as Rodrigo suggests (i'd never noticed this command 
before), but it's not obvious to me what the score variable that -_robust- 
requires should be.

I'd be curious to know a bit more about your application, in particular 
what your clusters are. In your example you specify varA as a covariate and 
also as the cluster id variable - i have difficulty imagining how this 
could be useful.

Best wishes,
Roger.

--On 28 June 2006 16:13 -0400 "Rodrigo A. Alfaro" <ralfaro76@hotmail.com> 
wrote:


> Erick,
>
> I don't know what metareg does. But maybe you can use -_robust- after the
> regression.
> There is a better description of -_robust- in Programming book.
>
> Rodrigo.
>
>
> ----- Original Message -----
> From: "Erick Turner" <turnere@ohsu.edu>
> To: <statalist@hsphsun2.harvard.edu>
> Sent: Wednesday, June 28, 2006 2:57 PM
> Subject: st: metareg w/ clustering?
>
>
> Is there a way to combine the cluster command with -metareg-, as can be
> done  for other types of regression?
> I tried
>
> > . metareg g  varA varB varC, robust cluster(varA)
> but I got
>
> > option robust not allowed
> > r(198);
> Tech support couldn't say much, since -metareg- is user-written, but
> suggested something might be done with regular regression but using
> analytic  weights.
>
> Thanks for any thoughts.
> Erick
*
*   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/