Post date: Sep 05, 2013 4:57:3 PM
Estimates of recombination are still lower between things on the same scaffold than on different scaffolds (and substantially so), but are also still to high (most loci on the same scaffold have r > 0.2). Moreover the recombination rate is not correlated with the distance between loci on the same scaffold. I suspect the failure to estimate recombination rates accurately (I assume that this is what is happening) is caused by uncertainty in offspring genotypes.
I revised the recomb source code to deal with this issue. Specifically, there is now a switch to enter a minimum level of certainty in the higher probability genotype for each individual and locus. Genotypes are sampled for these high certainty individuals, but not for others, and thus recombination rate estimates are based solely on these higher certainty genotypes. I added the number of high certainty genotypes that is the fifth value in the output for each line. I am running the new version of recomb on the dorc cluster with minimum certainty of 0.8 or 0.9. This is all with family 25266x25267.