halasadi
commented
9 years ago We've implemented this correction and it works well! See: https://github.com/halasadi/ancient-structure/blob/master/docs/issue10.pdf
Closed halasadi closed 9 years ago
halasadi
commented
9 years ago We've implemented this correction and it works well! See: https://github.com/halasadi/ancient-structure/blob/master/docs/issue10.pdf
Small samples (as typical with ancient DNA) give noisy estimates of the ancient allele frequencies. So we'd like to incorporate a fixed effect error to accommodate small sample sizes. The error is roughly f(1-f)/n where f is the frequency and n is the sample size. Since it is fixed effect, we can still use a similar EM update as Skotte et al 2013.
In our case, the Loscbour ancestral allele frequency is estimated from only one individual so we know that this is a major concern in our analysis.