Hi everybody. I uploaded a new version of the "felogit" package for conditional ML estimation of fixed-effects panel logit models (nothing revolutionary, really: it's just Chamberlain (1980)). I changed a few internals and the new code should be considerably faster than the old one. From my experiments, felogit() is about 3 times faster than the equivalent stata command (xtlogit with the fe option). However, a few robustness checks wouldn't hurt 😉. So, if you have a (preferably large) panel dataset to run a few checks, I'd appreciate your feedback. Thanks! ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------