Diary And Notes For: Jari Oksanen
nestedtemp: test results
Posted on 2008-02-14 20:35
As I wrote in my message of yesterday ("Analysis of nestedness: matrix temperature"), I am still not sure whether I should include matrix temperature calculations of community nestedness in vegan. Today I had some time to check the results of the vegan function against values reported in the literature. I run a couple of tests against Atmar & Patterson (1993). As I expected, results were not identical: I really do not know how to reproduce their software. However, the three example data sets in the paper gave very similar results: Small mammals of Rocky Mountains: A&P 3.81, vegan 3.59 Baja land birds: A&P 7.25, vegan 7.75 Queen Charlotte Island land birds: A&P 15.87, vegan 14.51 Differences probably are mainly in defining the "smooth" fill line, and possibly in defining the diagonal distances through surprises, in particular in the corners. However, the results are so similar to the published standards that I may have nestedtemp in vegan (with a warning to the user). SOURCES & REFERENCES Atmar & Patterson paper may be easiest to access through http://aics-research.com/research/order.html The analysed matrices were available in their binary software package: http://aics-research.com/nestedness/tempcalc.html This is a Windows binary file, but you can simply unzip the exe and you get a directory ("folder") called MATRICES which contains example data sets in plain ascii. The first lines till the line "endnotes" give a brief description of the data and the literature references. I could extract data in all these cases with the following dirty and inelegant piece of R: nestedRead <-function (file) { d <- readLines(con = file, warn = FALSE) endnotes <- which(d=="endnotes") d <- d[(endnotes+1): (length(d)-1)] d <- d[sapply(d, nchar) > 0] d <- strsplit(d, "") sapply(d, as.numeric) } This would be much more beautiful in Python. This was tested only with a couple of the files, and it may not work with all idiosyncratic cases.
