Reader Comments

Post a new comment on this article

Error in likelihood ratio test

Posted by richfrommich on 25 Oct 2010 at 13:00 GMT

There is an error in our use of the likelihood ratio test for rejecting Model 1 (site-independent equilibrium frequencies) in favour of Model 2 (site-dependent equilibrium frequencies). In the paper, we incorrectly assumed that the degrees of freedom at each location was equal to the number of non-zero equilibrium frequencies minus one, that is, not including the equilibrium frequencies for the amino acids not found at that location. We should have considered all 19 degrees of freedom per site. When we repeated the calculation, considering 19 degrees of freedom for each site times the number of sites, we found we can still reject Model 1 in favour of Model 2. We would like to thank Simon Whelan and Konrad Scheffler for pointing this error out to us.

Competing interests declared: Author of this manuscript.