Re: st: RE: Problem with METAN (forest plot)...

[RJ Harris <ross.harris@bristol.ac.uk>]

Dear Julien

Sorry to be so tardy with my reply.

If you want odds ratios on the log scale there is an option for it (log), 
like this:

use http://fmwww.bc.edu/repec/bocode/m/metan_example_data
metan tdeath tnodeath cdeath cnodeath, or log

If you are pooling pre-calculated odds ratios, these must be converted to 
the log scale before your pool them:

gen logOR = ln(OR) ...etc.
metan logOR logORlci logORuci

Results will of course be on the log scale, unless you use eform.

Best wishes,
Ross


--On 27 June 2008 02:33 -0400 statalist-digest 
<owner-statalist@hsphsun2.harvard.edu> wrote:


> Date: Thu, 26 Jun 2008 10:38:15 -0400
> From: "Austin Nichols" <austinnichols@gmail.com>
> Subject: Re: st: RE: Problem with METAN (forest plot)...
>
> Julien--
> You can just make your own graph using the variables created by -metan-:
>
>     By default, metan adds the following new variables to the dataset:
>
>         _ES        Effect size (ES)
>         _seES      Standard error of ES
>         _selogES   Standard error of log ES
>         _LCI       Lower confidence limit for ES
>         _UCI       Upper confidence limit for ES
>         _WT        Study percentage weight
>         _SS        Study sample size
>
> and use the -xscale(log)- option and any others you like in the graph
> command.
>
> E.g.
> use http://fmwww.bc.edu/repec/bocode/m/metan_example_data, clear
> replace tmean=tm+10
> metan tsample tmean tsd csample cmean csd, by(type_study)
> gsort type_study -id
> g nid=_n
> loc o "leg(off) yti(" ") yli(10.5) barw(.01) horiz color(gray white) yla("
> forv i=1/20 {
>   loc o `"`o' `i' "`=id[`i']'" "'
>  }
> loc o `"`o' , angle(0))"'
> tw spike _UCI _LCI nid, `o' || sc nid _ES, name(notlog)
> tw spike _UCI _LCI nid, xsc(log) `o' || sc nid _ES, xsc(log)
>
>
> On Thu, Jun 26, 2008 at 6:50 AM, Nick Cox <n.j.cox@durham.ac.uk> wrote:
> > I have not tried it but it seems that -metan- (from SJ 8(1)) should
> > support -xsc(log)-.
> >
> > However, all that would do is take the graph that would be displayed and
> > change that one detail. No calculations would be carried out on a log
> > scale.
> >
> > Nick
> > n.j.cox@durham.ac.uk
> >
> > Julien D.
> >
> > I would like to make a forest plot to graphicaly present results from
> > stratified analyses with different subgroups (men, women, diabetics, no
> > diabetics...)
> > I use the METAN module with 3 cols table (OR, lower and upper confidence
> >
> > interval) but I am embarass with large confidence interval (1.7-52.3 for
> >
> > example) and would like to use a log scale for axis...
> *
> *   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/


----------------------
Ross Harris
Department of Social Medicine
University of Bristol
+44 (0)117 928 7338
Ross.Harris@bristol.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/