Re: st: meta-regression coefficient - adjusting?

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

--On 01 December 2004 14:31 -0500 ronald eysenck <reysenck@lycos.com> wrote:

> I am performing a meta-regression analysis using 'metareg.' When no
> covariates are in the model, the intercept should be the exact equivalent
> of a random-effect pooled estimate calculated with 'meta.'
>
> Now, the purpose of these models are: [1] to look at predictors of the
> effect estimate (or in other words, to ceck if any study-feature makes
> studies more likely to show a bigger or smaller effect size); [2] ...here
> is my doubt: how should we interpret the intercept in a metaregression
> model with covariates? Should it be considered as a pooled estimate
> adjusted for study-features, the same way we would adjust for confounders,
> say, in a logistic regression model? I have been taking a look in books and
> articles on the subject but have not been able so far to come up with a
> reasonably justified answer.
--On 03 December 2004 10:14 +0100 "G. ter Riet" <G.terRiet@amc.uva.nl> 
replied:


> My understanding is that the intercept in a model with covariates
> represents the pooled estimate in the subgroup in which all covariates
> have the value of zero. I do not know a reference on that.
I'd agree with that. Metaregression examines how the effect varies according 
to the covariate, it doesn't adjust for the covariate. This is no different 
from ordinary regression - by regressing y on x you're examining how y varies 
with x, not adjusting y for x. If y does vary with x it is not very useful to 
give a single value summarising y for all values of x. The intercept is 
simply the fitted value of y when x=0 and so isn't of any greater interest 
than the fitted value of y at any other value of x. That's why the intercept 
is seldom the focus of interpretation.

A good reference on how meta-regression analysis should be conducted and 
interpreted is:

SG Thompson and JPT Higgins, "How should meta-regression analyses be 
undertaken and interpreted?",Stat Med 2002;21(11):1559-1573

Roger.

--
Roger Harbord
Department of Social Medicine, University of Bristol, UK
http://www.epi.bris.ac.uk


*
*   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/