Hello, I have a couple questions regarding generalized linear mixed models specifically around fitting the random effects terms correctly to account for any pseudo-replication. I am reading through and trying to follow examples from Zuur et al. Mixed Effects Models and Extensions in Ecology with R, but am still at bit unsure if I am specifying the models correctly. Background information: Our study design: We used mark-recapture trapping to obtain density estimates and demography of prairie dogs. The location of our study sites are nested, we have 6 trapping plots within 3 colonies of prairie dogs. Within each colony 3 plots are treatment (given supplemental food) and 3 are control sites. Therefore, in total we have 18 study plots (9 control and 9 treatment). Our study spans over two years and within each year we trapped prairie dogs twice (Spring 2008, Summer 2008, Spring 2009, Summer 2009) for a total of four trapping sessions. Therefore we also have temporal repeated measures in our sampling design. From the mark-recapture data we estimated density using Program DENSITY for each study plot and for each session (a total of 4 density estimates for each 18 plots over time). We also did vegetation sampling to obtain estimates of the natural food available in each plot, which gave us total biomass and % edible vegetation as covariates. We are using the lme4 package in program R (2.11.1) for analysis: We fitted a GLMM with Laplace restricted maximum likelihood estimation. Since our response variable density is a count with overdispersion we used a corrected poisson distribution with a log link function. And in order to account for the nested and repeated measures in our design we want to include Plot nested in Colony as random effects. The explanatory variables include Treatment, Session, Total biomass, and/or %Edible food for the fixed effects. (Colony, Plot, Treatment, and Session are all factor variables). And an example of the R code that we started with is: PDmodel_1 <- lmer(Density~1+Treatment+Session+Biomass+(1|Colony/Plot), family=quasipoisson, data=DensityPD) My main question is: Does this accurately take into account our nested and repeated measures design? Have we accounted for any possible pseudoreplication with this type of analysis? Do I need to also look at type AR-1 error structures (or can I do that in a glmm?) Or is there a simpler way you might suggest? I can also look at the change in density between time intervals instead density at time intervals if that makes anything easier to analyze? I can also split Session into season and year if that is helpful too. Any help, advice or guidance would be greatly appreciated. And if you need any more information or clarification, please ask. Thanks so much for your time and any advice in advance, natasha -- View this message in context: http://r.789695.n4.nabble.com/GLMM-random-effects-tp2330558p2330558.html Sent from the R help mailing list archive at Nabble.com.
______________________________________________ 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.
