st: Metan - possible to amend the ado to use difference confidence intervals?

["Paul O'Brien" <paul.obrien@afe2.org.uk>]

Having completed the meta-analysis of randomised trials of  
intrauterine devices, I not want to combine all the individual results  
in a graph where I input individual pregnancy rates and the standard  
errors rather than risk difference estimates and their standard errors.

This is _not_ what metan is intended for, but it does what I want well  
- produces the graphs I am looking for.

however, the confidence intervals calculated by metan gives impossible  
values at the lower end when applied to individual rates - negative  
rates.

For the confidence intervals I am using: p ^ exp(+/- 1.96s/(p x  
ln(p))) as recommended by Kalbfleisch, J.D. and Prentice R.L. (1980).   
The Statistical Analysis of Failure Time Data. I actually use p x 100  
^ exp(+/- 1.96s/(p x ln(p/100))) as I have the data in percentages.

Would it be possible to change the metan ado to produce these  
confidence intervals? I also want to produce summary rates, and tests  
for heterogeneity.

Unfortunately, I have no experience of programming in Stata. I have  
looked at the metan.ado but cannot see where an amendment might go.

I have lots of graphs of this sort to make with lost of data, so to be  
able to automate it using metan would make a big difference to me.

Thanks for any help.

Paul


-------------------------------
Dr Paul O'Brien
Raymede Clinic
Westside Contraceptive Services
Westminster Primary Care Trust
Exmoor St
London
W10 6DZ
UK
44 + (0)20 8962 4450






*
*   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/