I am involved in the analysis of non-nested and nested data which can be modelled using bivariate probit or logit models. The data emanate from an economic setting. It would seem that there are two appoaches, (1) that of, for example, Heise and Myers (Biometrics, 1996) which uses c.d.f.'s for the appropriate bivariate distribution to calculate the requisite probabilities and (2) that of Glonek and McCullagh (JRSSB, 1995) which is based on log-linear models with a composite link. I have two questions!
1. Are the two approaches equivalent?
2. Is there any software available for either of the approaches? My colleague in Agricultural Economics has heard of a package LIMDEP which he thinks will handle approach (1).
Any comments would be most welcome.
Linda Haines
Professor Linda M. Haines
Program Director in Statistics and Biometry,
University of Natal Pietermaritzburg,
Private Bag X01,
Scottsville 3209,
South Africa
Tel: 033 2605610
Fax: 033 260 5648
Cell: 0824938326
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|