1

I have a response variable of percent cover of vegetation in a quadrat. I have tried to arcsine squareroot this data as recommended in the Crawley R book but I am not getting good fit. The data is zero-inflated with about half the data points being zero. Can someone point me in the right direction for how to approach building this model? I am striking out on Google.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
user1658170
  • 151
  • 1
  • 6
  • 1
    Not clear what model you seek here as you don't make it clear what predictors you may or may not be using. But in general if there's a spike in a distribution, there is a corresponding spike in the distribution of any transformation of it; that's inescapable. There are models for zero inflation of various response distributions; here you might want zero inflated beta. – Nick Cox Nov 25 '13 at 17:21
  • Regarding the arcsine transformation, perhaps you could read the paper [Arcsine is asinine](http://www.esajournals.org/doi/abs/10.1890/10-0340.1) before continuing this line of investigation? A free PDF version is [here](http://www.maths.unsw.edu.au/sites/default/files/wartonhui10-0340final_0.pdf). Also, this previous [stats.se] [discussion](http://stats.stackexchange.com/q/20772/1390) would seem of relevance: – Gavin Simpson Nov 25 '13 at 17:27
  • As for R software for zero-inflated beta, try the gamlss package. Also take a look at the VGAM package as it may be able to do this as well. – Gavin Simpson Nov 25 '13 at 18:07
  • You may try the pscl package in R. – overwhelmed Mar 20 '14 at 09:34

0 Answers0