Re: [R] R-help Digest, Vol 171, Issue 20
Thanks Ron, In fact, I want to make a model choice using different fixed structures and using the results of: Gurka MJ (2006) Selecting the best linear mixed model under reml. The American Statistician 60(1):19{26, the best criterium uses the reml likelihood. I asked the ASREML-r developpers and they answered that their results were checked against GENSTAT. I think it is not really a good think for the R community to compute a REML likelihood that is probably not the REML likelihood. Brigitte Brigitte Mangin, INRA, LIPM, CS 52627, 31326 CASTANET-TOLOSAN tel: 33 + (0)5 61 28 54 58 De : Crump, Ron <r.e.cr...@warwick.ac.uk> Envoyé : mardi 23 mai 2017 10:29 À : r-help@r-project.org; Brigitte Mangin Objet : Re: R-help Digest, Vol 171, Issue 20 Hi Brigitte, >Did somebody know why asreml does not provide the same REML loglikehood >as coxme, lme4 or lmne. I don't know the answer to this, but I'd guess it is either to do with the use of the average information REML algorithm or asreml-r is for some reason ending up with a different subset of the data. >If it was just a constant value between the two models (with or without >the fixed effect) it would not be important. But it is not. >I checked that the variance component estimators were equal. I'm still not clear that it is important (if the data subset analysed is the same). You would only use the REML likelihoods to compare models with different random effects and the same fixed effect structure (is there another use for the REML likelihood other than that?), so then it is really a question of whether for a given pair of random effect models and the same data the likelihood ratio test statistic changes across analysis methods. Unless for some reason you are comparing two random effect models fitted with different routines (one of which is asreml-r). Ron. __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see 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.
Re: [R] mixed Model: asreml-r versus nmle,lme4 or coxme
Thank's Thierry, but as i mentioned, it is not a constant depending only of the data, since with the same observed trait: the difference (between asreml and R packages) is equal to 29.40 in the model with a fixed effect (Type) and the difference is equal to 32.16 in the model with only mu. And that, it is a big concern. De : Thierry Onkelinx <thierry.onkel...@inbo.be> Envoy� : vendredi 19 mai 2017 16:40 � : Brigitte Mangin Cc : r-h...@lists.r-project.org Objet : Re: [R] mixed Model: asreml-r versus nmle,lme4 or coxme Dear Brigitte, Maybe because the log likelihood is calculated differently. Note that the log likelihood contains a constant which only depends on the data. So one can safely omit that part for model comparison, assuming that use you the same formula to calculate the likelihood for all models. Best regards, ir. Thierry Onkelinx Instituut voor natuur- en bosonderzoek / Research Institute for Nature and Forest team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance Kliniekstraat 25 1070 Anderlecht Belgium To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher The plural of anecdote is not data. ~ Roger Brinner The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey 2017-05-19 14:30 GMT+02:00 Brigitte Mangin <brigitte.man...@inra.fr<mailto:brigitte.man...@inra.fr>>: Hi, Did somebody know why asreml does not provide the same REML loglikehood as coxme, lme4 or lmne. Here is a simple example showing the differences: ### library(lme4) library(coxme) library(asreml) library(nlme) data(ergoStool, package="nlme") # use a data set from nlme fit1 <- lmekin(effort ~ Type+(1|Subject), data=ergoStool,method="REML") fit1$loglik #-60.56539 fit2 <- lmer(effort ~ Type+(1|Subject), data=ergoStool,REML=TRUE) logLik(fit2) #'log Lik.' -60.56539 (df=6) fit3<-asreml(fixed=effort ~ Type,random=~Subject,data=ergoStool, na.method.X="omit",na.method.Y="omit") fit3$loglik #-31.15936 fit4<-lme(effort ~ Type,random=~1|Subject, data = ergoStool,method="REML") fit4$logLik #-60.56539 fit1 <- lmekin(effort ~ (1|Subject), data=ergoStool,method="REML") fit1$loglik #-78.91898 fit2 <- lmer(effort ~ (1|Subject), data=ergoStool,REML=TRUE) logLik(fit2) #'log Lik.' -78.91898 (df=3) fit3<-asreml(fixed=effort ~ 1,random=~Subject,data=ergoStool, na.method.X="omit",na.method.Y="omit") fit3$loglik #-46.75614 fit4<-lme(effort ~ 1,random=~1|Subject, data = ergoStool,method="REML") fit4$logLik #-78.91898 If it was just a constant value between the two models (with or without the fixed effect) it would not be important. But it is not. I checked that the variance component estimators were equal. Thanks [[alternative HTML version deleted]] __ R-help@r-project.org<mailto:R-help@r-project.org> mailing list -- To UNSUBSCRIBE and more, see 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. [[alternative HTML version deleted]] __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see 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.
[R] mixed Model: asreml-r versus nmle,lme4 or coxme
Hi, Did somebody know why asreml does not provide the same REML loglikehood as coxme, lme4 or lmne. Here is a simple example showing the differences: ### library(lme4) library(coxme) library(asreml) library(nlme) data(ergoStool, package="nlme") # use a data set from nlme fit1 <- lmekin(effort ~ Type+(1|Subject), data=ergoStool,method="REML") fit1$loglik #-60.56539 fit2 <- lmer(effort ~ Type+(1|Subject), data=ergoStool,REML=TRUE) logLik(fit2) #'log Lik.' -60.56539 (df=6) fit3<-asreml(fixed=effort ~ Type,random=~Subject,data=ergoStool, na.method.X="omit",na.method.Y="omit") fit3$loglik #-31.15936 fit4<-lme(effort ~ Type,random=~1|Subject, data = ergoStool,method="REML") fit4$logLik #-60.56539 fit1 <- lmekin(effort ~ (1|Subject), data=ergoStool,method="REML") fit1$loglik #-78.91898 fit2 <- lmer(effort ~ (1|Subject), data=ergoStool,REML=TRUE) logLik(fit2) #'log Lik.' -78.91898 (df=3) fit3<-asreml(fixed=effort ~ 1,random=~Subject,data=ergoStool, na.method.X="omit",na.method.Y="omit") fit3$loglik #-46.75614 fit4<-lme(effort ~ 1,random=~1|Subject, data = ergoStool,method="REML") fit4$logLik #-78.91898 If it was just a constant value between the two models (with or without the fixed effect) it would not be important. But it is not. I checked that the variance component estimators were equal. Thanks [[alternative HTML version deleted]] __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see 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.