Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


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

RE: st: meta-regression - fitting quadratics


From   "Tiago V. Pereira" <tiago.pereira@mbe.bio.br>
To   dhjd2@medschl.cam.ac.uk
Subject   RE: st: meta-regression - fitting quadratics
Date   Mon, 19 Mar 2012 11:31:12 -0300 (BRT)

Dear Daniel,

I believe that there is no ready-to-use Stata package for your purpose. I
believe you would need to implement your (non-linear) meta-regression
model with a random intercept. Not sure how feasible that is in Stata.

Furthermore, prevalences may not have a normal distribution. So approaches
that transform them to approximately normal variables are particularly
required.

Based on what I have read from literature, 99.9% of the meta-analyses
combining prevalences/proportions as effect size display significant
statistical heterogeneity.

 If that finding reflect real clinical/methodological heterogeneity or bad
properties of that effect size, it remains to be investigated.

Perhaps a start point would be:

1: Baker R, Jackson D. Inference for meta-analysis with a suspected temporal
trend. Biom J. 2010 Aug;52(4):538-51. PubMed PMID: 20740482.


All the best,

Tiago





--
Dear Statalisters,

I'm wondering how to fit nonlinear meta-regression models.

I'm conducting a literature-based meta-analysis, combining estimates of
prevalence of drug resistance. Each study reports a prevalence and
(within-study) standard errors can be derived. The appropriate approach
for pooling these estimates is to use a random-effects model. One key
covariate is duration exposed to the drug (to which resistance has
developed). Using metareg, I can see how the prevalence estimates vary
as a function of duration. So far so good.

What are the appropriate post-estimation procedures that may (or may
not) point to fitting polynomial quantities? A priori, the relationship
between drug resistance and duration of treatment is *not* likely to be
constant over time, e.g. resistance might emerge more rapidly during the
first year or two, and then the relationship might plateau.

I'd be very grateful for your collective thoughts on how to approach
this.

Many thanks and best wishes,
Daniel Davis
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index