Hi everyone, I've searched the internet and lots of stats books high and low for this one, but nothing seems to be quite what I want. I've got continuous data on four different state activities recorded in seconds, however each continuous session is not equally long, so the data are best expressed as proportions, i.e.: Subject 1: Swimming, 0.5, Hiding, 0.25, Edge, 0.125, Inactive, 0.125 and so on for each subject. I have two fixed effects and a possible random effect that may only have a very small influence on the end results. The four response variables are obviously not independent as they are a proportion of the total. The data are not normal, even when arcsine transformed from proportions where only the response variable with the highest proportion comes close (that being swimming).
I have used four separate binomial glms with each response variable (in seconds rather than as a proportion) against total time, but I'd rather incorporate them all into one model somehow. I thought about doing a manova, but I'm pretty sure lots of assumptions aren't met. For one, the response variables aren't independent. It confuses me that it's so hard for me to analyse this data, as I assumed it would be relatively straight forward. Any ideas? Best regards, Freya [[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.