I've been using Stata commands glm and bprobit to look at dose-response curves, where response is r/n. The Stata Reference Manual under glm comments that "the only difference ... is how they went about obtaining the answer, although this difference is hidden from us. bprobit secretly expands the data to obtain [sum of n] observations so that it can run standard, individual level commands."
With respect, it's not hidden or secret, as bprobit shows the number of observations as the sum, not the number of groups. Although the fitted parameters and SEs are the same for the two models, a different log-likelihood value is reported. The glm fit is assessed by deviance, AIC and BIC; probit reports a LR chi-squared and pseudo R-squared.
I'd appreciate comments on which analysis colleagues would consider more useful, valid or intelligible to report. As a rider, on the offchance, does anyone have Stata code for back-predicting the dose with CIs?
Allan Reese
***********************************************************************************
This email and any attachments are intended for the named recipient only. Its unauthorised use, distribution, disclosure, storage or copying is not permitted. If you have received it in error, please destroy all copies and notify the sender. In messages of a non-business nature, the views and opinions expressed are the author's own and do not necessarily reflect those of the organisation from which it is sent. All emails may be subject to monitoring.
***********************************************************************************
|