A standard method for determining the proportions of salmon originating
from different populations in a mixture harvested in the ocean uses allele
frequencies observed in the source populations and the observed multilocus
phenotype frequencies in the mixture to make a maximum likelihood estimate
of the mixture proportions. During the two months of my funding, I initiated
a project familiarizing myself with previous work and investigating the
information available in genetic data for this sort of mixture deconvolution.
In particular, I investigated the Fisher information for estimating the
mixture proportions when considering only allele frequencies or single-locus
phenotype frequencies in the mixture, as against the Fisher information
when considering multilocus phenotypes (the sufficient statistic). These
preliminary investigations led to a current project which involves estimating
the mixture proportions via a Bayesian approach, using reversible-jump
Markov chain Monte Carlo. |
