[R] metaplot and meta.summaries

[Thomas Lumley]

On Thu, 24 Aug 2006, Anne Katrin Heinrichs wrote:
> metaplot:
> ---------
>
> Can I change the label size? I've got 126 values and the intersection of the labels makes
> it impossible to read them.

Yes. metaplot() accepts the cex graphics parameter. This doesn't alter the 
size of the summary text, so you might want to have no summlabel= and add 
it on later.

> Why do I have to give sumse (Standard Error) and sumnn (Precision) of 
> the summary estimate? I can calculate one from the other, right? Just to 
> make sure I'm not misunderstanding something.

It's an oversight by the designer.  metaplot() was basically designed to 
be called from the plot methods for the meta. objects.

You might also want to look at forestplot(), which is more flexible. I 
don't know how it copes with large numbers of intervals, but if it doesn't 
it would be worth fixing.

> metaplot(CoeffVector, StdErrorVector, nn=NULL, labels=Name, conf.level=0.95,
> xlab=paste(CoeffNames[j], CoefficientName[i]),
> ylab="Countries",
> xlim=NULL, summn=PostCoeffs[1,j], sumse=sqrt(PostVars[1,j]),
> sumnn=1/(PostVars[1,j]),
> summlabel="Summary",
> lwd=2, boxsize=1)
>
>
> meta.summaries
> --------------
>
> What does it mean, when I get the warning message that NaNs were produced in "pchisq(q,
> df, lower.tail, log.p)"? Is there something wrong with my data (there are no NAs in the
> data)?
>
> MetaAnalyse <- meta.summaries(CoeffVector,
> 					StdErrorVector,
> 					method = "random")
>

It means that NaNs were produced.  You should be able to see where from 
the output. It's hard to say more without more information.

 	-thomas

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle