At 01:11 20/07/2010, Kate Snedeker wrote:>Greetings
>
>I have been trying to use metaplot to create a forest plot of my
>meta-analysis results. My data is in the form of ORs, so I've been
>inputting the ORs as the point estimate (mn) and the SEs of the
>logORs as the SE (se). However, whilst the point estimates appear
>to be displayed properly, many of the confidence intervals go below
>zero (not possible for an OR, as far as I know).
Kate, did you really mean that you are entering the OR and the se of
logOR? If so then I imagine using logOR and the se of logOR would be
a big advance.
>I think the confidence intervals shown might be simply 1.96 x SE or
>2 x SE. Is there a way to get meta plot to show the proper 95% CIs
>for the ORs?
>
>I have tried forestplot, but am having a lot of issues trying to get
>the columns to line up correctly, and in labeltext, it doesn't see
>to want to read the values I'm listing for the names of the studies.
>All I get are a changing list of numbers.
Posting an example here might help.
>[As a note, since the studies I'm meta-analysing use multi-level
>regression, I can't simply input the raw numbers for
>successes/failures since the ORs are all adjusted for various factors.]
>
>Thanks!
>Kate Snedeker
>
>--
>Kate Snedeker, PhD
>Postdoctoral Fellow
>Centre for Public Health and Zoonoses & Department of Population
Medicine
>MacNabb House
>Ontario Veterinary College
>University of Guelph
>Guelph, Ontario N1G 2W1
>
>Phone: 519-824-4120 x54471
>Fax: 519-766-1730
>E-mail: snedeker at uoguelph.ca
Michael Dewey
http://www.aghmed.fsnet.co.uk