Hello,
I'm running BayesAss on a dataset with 6 loci and 9 putative populations, currently using 6,000,000 generations plus a 1,500,000 burn-in; I've got two questions I hope you can help with.
1) The geographic distance between adjacent sample sites is substantially greater than actual potential dispersal distances for individuals over 1 or 2 generations. Is the program inappropriate for this type of sampling?
2) L(X/S; M, t, F, p) converges quickly and likelihood scores are reasonably consistent across runs. However, L(M, t/m) fluctuates considerably throughout the course of a run, varies greatly across runs, with some runs having positive scores (how is this possible?). Presumably as a consequence, source populations for immigrants are inconsistent across runs. Within a run, one or two populations are identified as major sources of immigrants into all other populations. I would love to fix this problem and get results I can have some confidence in but I suspect that there's not enough information in my dataset to do so. I'd be really grateful if you could tell me if I'm right, and perhaps give me some idea of how to interpret these results in terms of what info is missing. For example, would I see this lack of consistency if no populations had exchanged migrants in the recent past?
Sorry for the long e-mail!
Cheers,
Polly
