On Tue, 13 Apr 2010, ONKELINX, Thierry wrote:
So your respons variable behaves like a continuous variable except that is range is limited to the 0-1 interval. In such a case I would transform the respons variable (e.g. logit, sqrt(arcsin())) and use a gaussian model.
A logit-Normal has variance roughly mu^2(1-mu)^2 and a quasibinomial logistic uses mu(1-mu), with the parameters having the same interpretations. The question of which variance function best approximates the data should really be an empirical one, not an a prioiri one. There is an example of exactly this in the quasilikelihood chapter in McCullagh and Nelder, where the observations are the proportion of damage on a set of leaves. -thomas Thomas Lumley Assoc. Professor, Biostatistics tlum...@u.washington.edu University of Washington, Seattle ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
