One place to start might be Frank Harrell's design library: http://hesweb1.med.virginia.edu/biostat/s/help/Design/html/Overview.html
There is a function called lrm for logistic regression models. I do not know what algorithm is used, however it does use MLE. You might get more information from the r-help mailing list. You can get instructions for joining through www.r-project.org. Good Luck, brett -----Original Message----- From: Ryan Schram [mailto:[EMAIL PROTECTED] Sent: Thursday, May 29, 2003 12:36 PM To: [EMAIL PROTECTED] Subject: BHHH Vs. IRLS Dear Statisticians: Apologies for cross-posting this (it was posted earlier to aus.stats.s). I am using R-1.7.0 for a logit analysis, as a part of a graduate seminar in quantitative methods. The rest of the class is using LIMDEP, which (I think) uses the Berndt, Hall, Hall, and Hausman (BHHH) algorithm for doing MLE. The glm package in Splus/R uses iteratively reweighted least squares (IRLS). What's the difference? Specifically, (1) Could someone point me to articles for a more social-science audience which discuss these algorithms? I'm especially interested in what conditions call for one or the other, or what are the strength and weaknesses of each.* (2) Is there a way to run each algorithm in Splus/R on the same model and see what happens? (3) Can I do this: > myfit <- glm(Y ~ X1 + X2, family=binomial(link=logit), trace=T) > plot( trace(myfit) ~ iterations(myfit) ) If not, how would I go about examining the estimation steps and what problems should I be looking for? Thanks for your help in advance, Ryan * I am using household survey data from a 1996 study of Papua New Guinea (N=1396). My models revolve around explaining living standards. This one, for instance, will test determinants of a household reporting a temporary food shortage. About 20% of those surveyed said they had a food shortage. ======================================================================== Ryan Schram Doctoral Student in Anthropology UC San Diego . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
