Hello, I have binomial data which seem to be best modelled with a link function between the logit and the complementary log-log. The logit works good for small p's but does not reach p=1 fast enough. The c-loglog reaches p=1 in a satisfactory manner but does not work as good as the logit for small p's. So I would like to use the asymmetric link function proposed by Aranda-Ordaz (1981):
g(p;a) = log(((1-p)**-a - 1)/a) I don't want to use this to check for the adequacy of the link function, I want to model my data with this link function. Hence, I need to estimate three parameters (a and the two parameters of the linear term: b + c*log(dose); I have only log(dose) as influencing variable.) So here is my question: Where do I find information about programming? or even a SAS MACRO? or an example which I can recalculate myself? I have found something which works in R and in GAUSS but I have no experience with these languages. I could learn R but actually I do not have enough time. Many thanks in advance, Claudia Nicolay Institute for Medical Biometry, Informatics, and Epidemiology Faculty of Medicine University of Bonn Germany . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
