Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: Meta analysis in single group,

From   "Tiago V. Pereira" <>
Subject   RE: st: Meta analysis in single group,
Date   Thu, 5 Jan 2012 14:36:09 -0200 (BRST)


If I am not wrong, within-study variances are usually assumed to be known,
but are estimated from the data. So, once the combined studies are of
approximately equal size, bias in the summary estimate is likely to be
small, if any.

The only problem I see for the `pre-pos' case is regarding the correlation
estimate between time points (i.e. the correlation between pre and pos) if
one does not have the raw data. Assumption of zero correlation will
provide a conservative Wald test (Z test) if the true correlation is >0,
but an anti-conservative Z test otherwise. Again, if the studies to be
combined are  approximately of equal size, bias in point estimates will be


What should he do about the possibility of substantial bias when
estimated variances are used in inverse-variance weighting and the
DerSimonian-Laird method?

David Hoaglin

Dear Asad,

I solved that problem writing my own code (both fixed-effects 'inverse
variance' and DerSimonian-Laird methods). They are easy to implement, and
I think that in your case this is the only way out.

-metan- has an option to perform analyses using the effect and its
standard error. However, that approach will not be suitable when data is
continuous, since the normal approximation may not be valid depending on
the sample size.

To address your question, you need  to estimate the mean difference (pos
vs pre) for each study. Then, an appropriate variance estimate should be
computed. To do that, a proper correlation estimate has to be known (or
estimated from the data).

Once you have the mean effect and its variance, a variance-weighted
analysis can be performed.

Let me know if you need help.


*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index