Hi,
thanks. I am indeed interested in the main effects of A and B and their interaction+ I want to incorporate C (the block or 'repetition' within which the A and B treatments were applied) as a random variable. So A*B would be the way, however errors of A and B are different due to different experimental plot sizes. When doing Anova the correct code should be this: summary(aov(ln_response) ~ A*B + Error(rep/A), data=Exp2) in which case the effect of A is calculated by using error A*rep and the effect of B and A*B is calculated using pooled error of B*rep and A*B*rep This I dont know how to specify in glmer. Maybe 'nesting' is not a right term to use (?) > To: r-h...@stat.math.ethz.ch > From: bbol...@gmail.com > Date: Fri, 18 Jan 2013 14:07:02 +0000 > Subject: Re: [R] Nesting fixed factors in lme4 package > > Martina Ozan <martina_ozan <at> hotmail.com> writes: > > > Hi, can anyone tell me how to nest two fixed factors using glmer in > > lme4? I have a split-plot design with two fixed factors - A (whole > > plot factor) and B (subplot factor), both with two levels. I want to > > do GLMM as I also want to include different plots as a random > > factor. But I am interested on the effect of A a B and their > > interaction on the response variable. I tried > > this:glmer(response~A*B+(A/B)+(1|C),data=Exp2,family=poisson but it > > gives the same output as if I removed (A/B) all together or used > > (A:B) instead thus the output is the same as: > > glmer(response~A*B+(1|C),data=Exp2,family=poisson anyone can help > > with how I define this nesting, so that data are analysed correctly > > given my split-plot design? thanks, Martina > > In general mixed model questions should go to > r-sig-mixed-mod...@r-project.org , but this is actually *not* > specifically a mixed model problem. If A and B are fixed factors, > you're typically interested in A*B, which translates to 1+A+B+A:B, > i.e. intercept; main effects of A and of B; and the interaction. > The nesting syntax A/B translates to 1 + A + A:B, i.e. no main > effect of B. Nesting would typically make more sense in a random-effects > context where the meaning of "B=1 in unit A=1" is different from > "B=1 in unit A=2", i.e. where you don't want or it doesn't make > sense to estimate a main effect of B across levels of A. > > Ben Bolker > > ______________________________________________ > 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. [[alternative HTML version deleted]] ______________________________________________ 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.