ML analysis (2)

Brice Quenoville quenovib at si.edu
Fri Jan 28 18:02:45 EST 2000


Hi,

I  realised my question on "branch and bound" search for ML analysis could be misunderstood because I guess I mistranslated my french word in english. When I said that the exhaustive search led to a different tree with a slighty higher -Ln L value, I meant a better.

Branch and bound: 6h45mn, score:10745.09295
Exhaustive search: 47 mn, score: 10534.83665
(for info) Heuristic search: 24 mn, score: 10745.09295 (same topology as Branch and bound)

So I don't have a problem with the fact that the exhaustive search gave me a better tree but with such different computation times and final results.

I also would like to add one question about the HKY tests for competing topology. When I do it, I have some cases where the HKY test reject the hypothesis of no differences between 2 trees whereas a simple LRT test (with nbr of taxa -2 as the df) gives me strong support for difference. I read that HKY test is no good when there is an a priori that a topology is better than another, because this violates one of its parameter assumption. As my 2 topologies come from exhaustive search of separate analysis of mit. and nucl. sequences data I guess I have a strong a priori. However, I see very often such test used in the literrature with no much details about assumptions. So I just would like some advice about the range of application and confidence of both test to compare two competing topologies.

Thank you very much again,
Brice

Brice Quenouille

Smithsonian Tropical Research Institute
Unit 0948
APO AA 34002-0948 USA
           or
Institute Smithsonian de Investigaciones Tropicales
Apartado Postal 2072
Balboa, Ancon
Republica de Panama



---







More information about the Mol-evol mailing list