On Wed, 23 Jul 2008, silvia narduzzi wrote:> Dear R users,> I use the glm() function to fit a generalized linear model with gamma > distribution function and log link. I have read in the help page that > the default method used by R is "glm.fit" (iteratively reweighted least > squares, IWLS). Is it possible to use maximum likelihood method?IRLS does implement maximum likelihood for the coefficients of the linear predictor. glm() does not estimate the gamma shape (summary.glm does, though), and MASS (the book and packages) discusses how to find the MLE of the shape. See library(MASS) ?gamma.shape> Silvia Narduzzi > Dipartimento di Epidemiologia > ASL RM E > Via di S. Costanza, 53 > 00198 Roma > Tel +39 06 83060461 > Mail narduzzi at asplazio.it-- 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
Dear R users, I use the glm() function to fit a generalized linear model with gamma distribution function and log link. I have read in the help page that the default method used by R is "glm.fit" (iteratively reweighted least squares, IWLS). Is it possible to use maximum likelihood method? Thanks Silvia Narduzzi Dipartimento di Epidemiologia ASL RM E Via di S. Costanza, 53 00198 Roma Tel +39 06 83060461 Mail narduzzi at asplazio.it
dear Silvia, quoting Venables WN and Ripley DB (1994) Modern Applied Statistics with S-plus, sringer, pag 185: "Since explicit expressions for the maximum likelihood estimators are not usually available estimates MUST be calculate iteratively" means that glm.fit performs MLE indirectly and efficiently. Hope it's help regards PF 2008/7/23 silvia narduzzi <narduzzi at asplazio.it>:> Dear R users, > I use the glm() function to fit a generalized linear model with gamma distribution function and log link. > I have read in the help page that the default method used by R is "glm.fit" (iteratively reweighted least squares, IWLS). > Is it possible to use maximum likelihood method? > > Thanks > > > > Silvia Narduzzi > Dipartimento di Epidemiologia > ASL RM E > Via di S. Costanza, 53 > 00198 Roma > Tel +39 06 83060461 > Mail narduzzi at asplazio.it > > ______________________________________________ > R-help at 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. >
silvia narduzzi wrote:> Dear R users, > I use the glm() function to fit a generalized linear model with gamma distribution function and log link. > I have read in the help page that the default method used by R is "glm.fit" (iteratively reweighted least squares, IWLS). > Is it possible to use maximum likelihood method? > >Same thing.... (Well, one is an algorithm, the other a principle, but in GLMs, successful application of IRLS leads to the maximum likelihood estimates.) -- O__ ---- Peter Dalgaard ?ster Farimagsgade 5, Entr.B c/ /'_ --- Dept. of Biostatistics PO Box 2099, 1014 Cph. K (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 ~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk) FAX: (+45) 35327907