Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Meta-analysis with 2 zero cells‏


From   Elliott Dasenbrook <edasenb1@jhmi.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: Meta-analysis with 2 zero cells‏
Date   Thu, 20 Nov 2008 13:03:08 -0500

In our analysis ONE study reported two Zero cell events: 0/n1events in the intervention group and 0/n2 events in the treatment group;

*The study is either excluded or creates very large confidence intervals...

Our codes is as follows:

metan  t_outcome t_nooutcome c_outcome c_nooutcome, rr fixedi...

Here is what we have tried so far:
1) When we run our usual metan(with both cells as Zero), this study is EXCLUDED 
2) When we run our metan and add a continuity correction  "cc(0.5)" , the study is excluded.
3) When we manually add 0.5 to the zero cells and use the command "nointeger", the confidence intervals span from 0 to 100 and present an unpleasing forest plot.
4) It didnt make a difference if we used inverse variance (fixedi) or Mantel and Haenszel (fixed)...

Most of the solutions I have found by searching seem to be if you have one zero cell;

We want to have this particular outcome be a figure in our manuscript, so it is important that the forest plot be displayed correctly.

Any guidance on suggestions/references for how to handle this would be greatly appreciated.

Thanks

Elliott
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index