[R] Help with nonlinear regression of binomial data

Harborne, Alastair A.R.Harborne at exeter.ac.uk
Fri Oct 24 12:48:32 CEST 2008

Hi all,
I have a manuscript nearly ready to go looking at a factorial analysis (inside and outside a Caribbean marine reserve) of how the biomass of predatory fish affects the proportion of survey transects containing urchins (more predators leads to fewer urchins). I would like to add some regression analysis, but my linear models with binomial errors are overdispersed because the data are nonlinear (negative power by the look of it). I have spent some time playing with gnlr but have realised that it is unlikely that I will be able to get the analysis to a publishable standard. If anyone is interested in working with me to complete this analysis I would happily include them as a co-author on the paper. The data are appended below in case anyone wants to see it.

Pred_biomass            proportion of transects with urchins
372.25                  0.28571
173.66                  0.14286
205.00                  0.28571
3436.77                 0.00000
862.54                  0.00000
387.82                  0.00000
385.50                  0.00000
345.83                  0.28571
132.61                  1.00000

- Note there seems to be an additional wrinkle as the biomass data are difficult to normalise.

Dr Alastair Harborne

Marine Spatial Ecology Lab
School of Biosciences
Hatherly Laboratory
University of Exeter, Prince of Wales Road
Exeter, EX4 4PS, UK

Phone: 01392-263784
Fax: 01392-263700
Skype name: al_harborne
Group website: http://www.ex.ac.uk/msel
Personal website: http://biosciences.exeter.ac.uk/staff/profiles/profile.php?id=alastair_harborne

More information about the R-help mailing list