Hello!

I usually run GARD on a local cluster to detect genome wide intragenic recombination. In the process I don't collapse identical sequences.

The other day someone in my lab did run a GARD analysis on datamonkey on a single gene that I had already analysed. Some sequences were identical, and so, probably to save computation time, datamonkey collapse the data to unique haplotypes (21 to 8 taxa). The "surprise" is that the results are not the same:
- with some duplicates, no recombination detected (c-AIC : 1284.35)
- with unique haplotypes, 1 breakpoint (c-AIC : 1219.91)


Well, I just don't know ho to interprate this, and which one we should believe. Any idea?


Thanks,

-Tristan

The delta c-AIC is 1.18956
The model parameters vary, but only slightly.

I'm not used to play with AIC improvement values, would you in that (and other cases) look at it before taking any decision? Do you use some sort of rule of thumb as with the Bayes factors?

-Tristan