I am working on some data based on a ratio from MacArthur and Wilson's Theory of Island Biogeography. If I take area of the source population as a constant, since all islands under consideration were colonized from the same source; and I take the taxon-specific variables as constants also, because I am comparing different islands' inherent probability of colonization by any given taxon, that leaves me with: (perpendicular diameter * e^-distance)/(2*pi*distance) The problem: when I use the raw data, I get infinitesimally small numbers, e.g. one island's calculation resulted in 1.24504e-315, and approximately half the islands ended up with zero. That cannot be right, because all the islands have been colonized by at least some species. So, going through Gotelli and Ellison, _A Primer of Ecological Statistics_, it appears I need to log transform the data. This seems to work; the same island above now has a colonization probability of 0.0178. But Gotelli and Ellison then go on to say that a result obtained from log transformed data should be back transformed in the final presentation. This gives a result of 1.0418. This cannot be, because there is no such thing as a probability >1. I cannot figure out where I went wrong. Jason Hernandez
