Hello,

I have a couple questions about using the KH test with GARD:

1)  In your paper, you use the SH to process your GARD results, but not in the single breakpoint analysis.  Do you suggest using the KH test to test phylogenetic congruence for the single breakpoint analysis as well?

2)  If you do suggest this, what is the best way to implement it?  I have run the batchfile KHTest.bf using the trees and breakpoint found when using SingleBreakpointRecomb.bf.  I get a p-value for 1st and 2nd partitions but am not sure how to interpret this result.  Does a significant p-value for either partition indicate recombination?  Or do both partitions have to have a significant p-value?

3)  In using the KH test to process GARD results as well as single breakpoint analysis results, is some correction for multiple testing implemented?

Thank you for your help!

Rachael

Hi Rachael,

To answer your questions

1). For the set of two trees (left and right of a breakpoint), KH and SH tests are equivalent.
2). GARD as implemented on Datamonkey.org already includes KH testing. Within HyPhy, GARDProcessor.bf (a standard analysis under 'Recombination') will also run KH testing. KHTest.bf will preform two tests: does tree 1 fit partition 1 better than tree 2, and does tree 2 fit partition 2 better than tree 1. If the maximum of these p-values is small (e.g. <0.05 or <0.025 etc), you can conclude that the topologies are discordant.
3). GARDProcessor.bf and datamonkey.org implement a simple Bonferroni MT correction.

Cheers,
Sergei