You might like to go to a full Bayesian smoother like Flexi distributed by
AgResearch, which I have used quite successfully as to fit does response
estimates. It will produce plots with confidence intervals of the resulting
curve, or you can obtain the estimates in tabular form. Likewise you can
ask for estimates at infinity and so get confidence intervals for the final
probalbilities. You can test for differences between groups by seeing
where, if anywhere, the confidence intervals don't overlap.
As an added bonus you don't have to assume any particular form for the
curve. Just that it has asymptotes. Flexi is a full GLM type smoother so
it fits binomial data semi-properly. (ie allows the variance of the
distribution to depend on the estimated mean. Not quite as good as Bugs
which fits the full distribution.)
The program runs on either a PC or VAX computers. Rod Ball at
[log in to unmask] has a unix version. The program is free but AgResearch
charge $40 for the cost of copying the 300 page manual. Electronic version
of the manual is free. Let me know if you want a copy.
Martin
Martin Upsdell Statistics Section
[log in to unmask] Ruakura Agricultural Research Centre
64-7-838 5149 Private Bag 3123, Hamilton
New Zealand
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|