The method in brief
The method computes all pairwise differences between all sequences from both samples. It does NOT count indels as differences unless they were recoded (i.e. only when '-' is compared to a '+', it counts as a difference). It then calculates the mean pairwise difference (Ks) and the mean pairwise LOG(1+Diff) (K*s) within each sample. Then, for each replicate, it randomizes the samples (which sequence in which sample) and recomputes those values. The probabilities are the fraction of random samples that have a smaller Ks (or K*s) than the real samples. For more details, please refer to the original paper by Hudson, Boos and Kaplan (1992).NEWS
A new version with a graphical output that can take several groups is available. Please visit: http://wwwabi.snv.jussieu.fr/public/mpweb/ it will replace soon this page. any suggestion/comment is more than welcome.
Online version of the test
This program has been written in C. It is originally part of a library (libdiversity) of functions that does manage aligned sequences and compute many standard population genetics metrics (The library is available upon request). The source of this stand-alone web-adapted version can be viewed at hudsontest.c (compiled as 'gcc -o hudsontest.cgi -O2 -DMACOSX hudsontest.c -lm'). I would be happy to incorporate any interesting suggestions.
If you have any suggestions/comments, please send them to Guillaume Achaz: firstname.lastname@example.org