Like you I also think that the main problem are not normally distributed variables and the eigenvalues encountered during processing, as reported by the program.
I gues the discontinuity in the variables value is mainly due to the small size of the sampling grid (50x50m), so most of the variables take the value zero (the EGV is not present within the square) or maximum (EGV covering the whole of the square), with very few intermediate cases. I had to choose a small size grid commensurate with the species home range. So I had to digitalize the environment variables (EGV) having the need to detect the same with much more detail than the coverage available (Corine Landcover, etc..). By using photo interpretation and field surveys I have detected hydrological variable (torrent, streams, flood areas, etc.), vegetation (meadows, forests, scrublands and mixed coenoses) and lithology / geo-morphology (rocky outcrops, debris, alluvial sediments, conoids, stony ground, etc.). Despite the detailed survey work, for the most part the distribution of EGV was not continuous enough (very far from the normal distribution)!
I will try to perform the analysis suggested to make more continuous variables (possibly asking to you some help on how to do).
Tanks to John and Diederik for your answers.
Paolo
Da: Biomapper-List@yahoogroups.com [mailto:Biomapper-List@yahoogroups.com] Per conto di Diederik Strubbe
Inviato: martedì 24 aprile 2012 9.23
A: Biomapper-List@yahoogroups.com
Cc: John Clark
Oggetto: Re: [BMp] Re: Discrepancies between theory (cross validation results) and graphical output (HSM)
Hey,
Indeed, I agree with John that it is rather suspicious that your first factor explains 100 % of the marginality and the specialization. My gues would be that there is a problem with the variables not being continuous enough, as continuous and normally distributed variables are one of the key requirements for an ENFA analysis. Can't you try to make them more continuous, for example by performing a roving window analysis on the non-continuous maps? Or use distance to some habitat features rather than their presence? Somewhere on the BIOMAPPER help page their is some information on converting (near) categorical data into continuous ones.
Hope this helps,
Diederik
On 4/23/2012 10:13 PM, John Clark wrote:
I wouldn't worry too much into the EGV being the inverse of suitability for the time being. That your first factor explains 100% of both marginality and specialization, and that 11 of 15 EGVs were not continuous enough and yielded large eigenvalues, indicate there may be a significant issue with the EGV's. You might want to take a closer look at them before focusing on mapping/validation. What sort of information were you interpreting from the photos? I'm by no means an expert, but unfortunately this list-serve is very quiet.
Best Regards,
John Clark
--
Dr.Diederik Strubbe
Evolutionary Ecology Group
Department of Biology
University of Antwerp
Middelheimcampus GV310
Groenenborgerlaan 171
2020 Antwerpen, Belgium
office: +32 3 265 3282
mobile phone: +32 477445568
skype user name: lakrinn
------------------------------
Biomapper FAQ: http://www.unil.ch/faq.html
Wiki-Biomapper: http://biomapper.wikispaces.com/
If you found this message useful, please consider copying it to the Wiki-Biomapper
------------------------------
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.