st: Re: R^2 after metareg is stata

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

Dear Anna,

R-squared for meta-regression isn't entirely straightforward as there is 
both within-study and between-study variation, so there's more than one 
possible R-squared-type measure. I think the most sensible one is the 
proportion of the between-study variance (tau2) accounted for by the 
covariate(s). This is like an adjusted R-squared. To calculate this you 
need to fit a model with no covariates as well as the model of interest 
with covariates and look at the values of tau2 in the output. Then

adjusted R2 = 1 - tau2(with covariates) / tau2(no covariates) 

Note that this will be 1 if tau2(with covariates) is zero.  As in OLS, its 
possible for adjusted R2 to be negative if the covariate(s) explain very 
little, as the estimate of tau2 can increase due to the degrees of freedom 
used up by the covariates.

I'll probably add this to the output of -metareg- in a future version.

Best wishes,
Roger.

--On 12 July 2006 14:41 +0200 "Matysiak, Anna" <Matysiak@demogr.mpg.de> 
wrote:


> Hallo
> Does anybody of you know how to compute R^2 after metareg is stata? I
> found your discussion in the stata faq, it seems that it is obvious how
> to do it, but I cannot find this option at all.
>
> Thanks a lot for your help in advance
> Anna
*
*   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/