Re: st: meta-analysis residual question

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

--On 17 February 2005 21:55 -0600 "Garrard, Wendy M." 
<wendy.garrard@vanderbilt.edu> wrote:


> Could someone help me understand the fitted values that are generated
> post-estimation using the metareg (updated version)?
>
> After a meta-regression I generated the fitted values using (predict
> yhat).
>
> But, I get quite different values when I use the regression equation to
> obtain the fitted values like so:
> 	(gen yhat2 = _b[_cons] + _b[Var1]*Var1 + _b[Var2]*Var2 -
> _b[Var3]*Var3)
>
> For example, for the first record in my data set, the fitted value is .5
> using the predict command, and 1.1 obtained via the equation.
>
> I assume this difference has something to do with how the weighting is
> used in the postestimation command, but I'd like to understand how to
> correctly approximate the fitted values using the equation.
Dear Wendy,

There's no reason for the two methods to differ. The weights aren't used when 
calculating fitted values (other than their effect on the coefficients 
already estimated). I can't reproduce this problem using my own example 
datasets. But I'm assuming the minus sign before the Var3 term in your code 
above is a typo for a plus - if not that could be the source of the problem!

If that's not it could you send me a full log-file using the procedure 
outlined at
http://www.stata.com/support/faqs/techsup/sendout.html
(please email me alone not Statalist and i'll summarise our finding to the 
list later)

Thanks,
Roger.

--
Roger Harbord
Department of Social Medicine, University of Bristol
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/