Hi, I am analysing an experiment that has one fixed (6 conditions) and two random factors (11 subjects, 24 images in the conditions). I read somewhere else that you can also see such a design as a nested experiment with the hierarchy: subjects -> condition -> image. For some analysis I have one respond variable and for others I have more. The response variables are non-normally distributed. Now the question: Is there a package that can deal with such a design? I would like to use a generalized linear model. Are there glms that are extended to do multivariate analysis (for the 2 random + 1 fixed variable design)? And how do you call such a design? Last question: Can you suggest me some literature about such a problem? I am quite unsure concerning the analysis. Thanks for any advice lisra [[alternative HTML version deleted]]
Interesting packages for you might be the nlme and lme4 packages and as a book Pinheiro/Bates, "Mixed-Effects Models in S and S-Plus" Lina Jansen schrieb:> Hi, > > I am analysing an experiment that has one fixed (6 conditions) and two > random factors (11 subjects, 24 images in the conditions). I read somewhere > else that you can also see such a design as a nested experiment with the > hierarchy: subjects -> condition -> image. For some analysis I have one > respond variable and for others I have more. The response variables are > non-normally distributed. Now the question: > > Is there a package that can deal with such a design? I would like to use a > generalized linear model. Are there glms that are extended to do > multivariate analysis (for the 2 random + 1 fixed variable design)? And how > do you call such a design? > > Last question: Can you suggest me some literature about such a problem? I am > quite unsure concerning the analysis. > > Thanks for any advice > lisra > > [[alternative HTML version deleted]] > > ______________________________________________ > R-help at stat.math.ethz.ch 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. > > >
Please always reply to the list as well as there always might be someone faster/better answering (or it could be that I am wrong, so someone might correct me) Indeed Pinheiro/Bates assume gaussian error terms... but I am not really sure whether you meant that with "non normally distributed respond variable" resp. "with non-normal data" however: "/ Mixed-effects models: / The recommended nlme <http://cran.r-project.org/src/contrib/Descriptions/nlme.html> package, associated with Pinheiro and Bates, / Mixed-Effects Models in S and S-PLUS / (Springer, 2000), fits linear and nonlinear mixed-effects models, commonly used in the social sciences for hierarchical and longitudinal data. Generalized linear mixed-effects models may be fit by the glmmPQL function in the MASS package, and by the lmer function in the Matrix <http://cran.r-project.org/src/contrib/Descriptions/Matrix.html> package (related to the lme4 <http://cran.r-project.org/src/contrib/Descriptions/lme4.html> package, which largely supersedes nlme <http://cran.r-project.org/src/contrib/Descriptions/nlme.html> for / linear / mixed models). Also see the lmeSplines <http://cran.r-project.org/src/contrib/Descriptions/lmeSplines.html> and lmm <http://cran.r-project.org/src/contrib/Descriptions/lmm.html> packages." [ http://cran.r-project.org/src/contrib/Views/SocialSciences.html ] Lina Jansen schrieb:> > > 2006/10/17, Stefan Grosse <singularitaet at gmx.net > <mailto:singularitaet at gmx.net>>: > > Interesting packages for you might be the nlme and lme4 packages > and as > a book Pinheiro/Bates, "Mixed-Effects Models in S and S-Plus" > > > Thank you for the answer. I am always unsure concerning the > non-normality. Can I use the nlme and lme4 with non-normal data? > First, I thought they would work like an ANOVA but with random and > fixed effects.