Matthew Sydes wrote:
> I am interested in performing tests of trend in ordered subgroups from a
> single randomised controlled trial. It is possible to use meta-analysis
> commands for this purpose, pretending that each subgroup is a separate
> trial. I have investigated the -meta- and -metan- commands. However,
> their tests for heterogeneity only seem to deal in global tests rather
> than tests for trends.
>
Like numbers-at-risk on KM plots, subgroup trend tests are an issue I often
wrangle with and find that I can solve only outside of Stata (albeit
somewhat unsatisfactorily and unloggable fashion).
Any help on how I may address this matter within Stata would be much
appreciated.
--------------------------------------------------------------------------------
Could you describe in more detail what it is that you're trying to do?
It seems that you have only aggregate-level data for the single study. Is the
problem that the outcome is continuous and the article shows only summary
statistics for the subgroups? If so, then take a look at -aovsum-, a user-
written command from John R. Gleason (-findit aovsum-). The help file for it
illustrates its use in combination with -anova- to "Test for linear trend in
dose, and for deviations from linear trend" in an example.
If the outcome is binary and the article provides counts (or proportions and
total sample size), then wouldn't conventional contingency table methods or
logistic regression work?
Joseph Coveney
*
* 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/
