0

I am trying to best analyse a set of foraging ecology data with >10 behaviour categories (DVs) and 3 levels of IV (season, sex, age). The time which an animal spent engaged in a behaviour was recorded and then divided by the total time spent in sight of the observer, so my data are proportional. As is typical, not all animals engaged in all behaviours and there are a large number of zeros in my dataset which is severely over-dispersed. I had initially analysed all the data in R using the glm function with family = quasibinomial, followed by anova. The intention was then to use the false discovery rate alpha to account for the large number of analyses. However, it has since been suggested that a multivariate approach might be better so I have been trying to figure out (a) if it's possible to run a quasi-binomial multivariate analysis of proportion data (b) how to go about it.

In the R documentation quasi-binomial family function page (from the VGAM package, the function mentioned above is quasibinomial() from base R) it is stated that if multivariate response = TRUE the response matrix should be binary. This seems a pretty straightforward indictment of my idea to run this analysis on my proportion data, but I am wondering why - is this just not possible and why not; or is there a particular package that could help?

Gavin Simpson
  • 37,567
  • 5
  • 110
  • 153
Mand
  • 1
  • 1
  • could you describe the statistical model more formally? – Xi'an Feb 08 '15 at 21:22
  • Hi, I want to run a MANOVA with 3 levels of IV on a highly overdispersed dataset where all values are bounded by 0 and 1. – Mand Feb 08 '15 at 21:51
  • If you really had multivariate binomial data, a multinomial model might have been appropriate; out of a number of trials, you have counts $y_i$ on a set of $m$ species. But you describe a proportion which I don't think fits into the multinomial model. A beta regression is often used for true proportions and dirichlet regression is a multivariate version of the univariate beta regression. There is an R package [**DirichletReg**](http://cran.r-project.org/web/packages/DirichletReg/index.html) which can fit those models. – Gavin Simpson Feb 08 '15 at 22:10
  • Question refers to a link, but none is visible. – Nick Cox Feb 09 '15 at 00:28
  • Assuming the OP is referring to the link in their post to the R-SIG-Ecology mailing list, they are linking to the `quasibinomialff()` family function from the **VGAM** package.. However, from what code they do show, they are using the `quasibinomial()` function from base R, which doesn't have the argument `mv` referred to by the OP. – Gavin Simpson Feb 09 '15 at 03:23

0 Answers0