 |
 |
In order to seek statistical consensus on best confidence interval methods and presentation of NNT statistics I asked for feedback on a web-based calculator at: http://194.205.65.119/statcalc/statcalc.htm Thanks to all who replied. The calculator has now been updated in response to suggestions from Jon Deeks, Robert Newcombe and Chris Cates. Full references are provided with the results given by the calculator. The presentation could be improved by providing guidance customised to levels of baseline risk. Routines for meta-analysis based on risk ratio, risk difference or odds ratio, with fixed or random effects, are written and validated - I may package these as web calculators in the future. Early experimental work includes stratified re-sampling in pursuit of 'bias-adjusted' pooled estimates. Any further feedback is welcome. Iain Buchan [log in to unmask] www.ukph.org/iain.htm p.s. apologies for cross-postings