Hello. I must fit a logistic regression to data in the form of proportions, but in which some of the proportions are zero. I therefore cannot use the glm function with a binomial link since the link function is not defined for p=0 or 1. What other solutions are available? Any references to this specific problem (i.e. regression using proportions, of which some are zero) would be welcome. Thanks. Bill Shipley [[alternative HTML version deleted]]
On Wed, 1 Mar 2006, Bill Shipley wrote:> Hello. I must fit a logistic regression to data in the form of > proportions, but in which some of the proportions are zero. I therefore > cannot use the glm function with a binomial link since the link function > is not defined for p=0 or 1.It is a logistic link, not a 'binomial link'. The fitted proportions cannot be zero (just 2*10^-16 which is exceedingly rare), but the observed ones can. MASS4 p.190 has an example with observed proportion zero.> What other solutions are available? Any > references to this specific problem (i.e. regression using proportions, > of which some are zero) would be welcome.Any good book on logistic regression.> Thanks. > > Bill Shipley > > [[alternative HTML version deleted]] > PLEASE do read the posting guide! http://www.R-project.org/posting-guide.htmlOh please do give up sending HTML mail. -- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595
Hello!>> Hello. I must fit a logistic regression to data in the form of >> proportions, but in which some of the proportions are zero. I therefore >> cannot use the glm function with a binomial link since the link function >> is not defined for p=0 or 1. > > It is a logistic link, not a 'binomial link'. The fitted proportions > cannot be zero (just 2*10^-16 which is exceedingly rare), but the observed > ones can. MASS4 p.190 has an example with observed proportion zero. >Yes, MASS4 p. 190 has example with 0 observed, but that is not a problem if you use a regression for dose. What can one do in case you have a level of a factor that has observed zero proportion or very low/high proportion?>> What other solutions are available? Any >> references to this specific problem (i.e. regression using proportions, >> of which some are zero) would be welcome. > > Any good book on logistic regression.-- Lep pozdrav / With regards, Gregor Gorjanc ---------------------------------------------------------------------- University of Ljubljana PhD student Biotechnical Faculty Zootechnical Department URI: http://www.bfro.uni-lj.si/MR/ggorjan Groblje 3 mail: gregor.gorjanc <at> bfro.uni-lj.si SI-1230 Domzale tel: +386 (0)1 72 17 861 Slovenia, Europe fax: +386 (0)1 72 17 888 ---------------------------------------------------------------------- "One must learn by doing the thing; for though you think you know it, you have no certainty until you try." Sophocles ~ 450 B.C.