This looks very relevant. A way to directly test how much a sampling location contributes to population structure.
A description posted to EvolDir by the author:
“I’d like to introduce a new tool for population genetics analyses called
ObStruct. ObStruct is software to objectively analyze ancestry profiles
produced by the Bayesian population genetics software packages Structure,
InStruct and BAPS. Presently, these ancestry profiles are visualized with
distruct and subjectively interpreted. ObStruct applies the R2 (R-squared)
statistic to objectively determine whether the fixed factor of interest
(e.g. geographic origin) correlates with inferred structure. It goes
on to determine the amount that each sampled and inferred population
contributes to the overall structure in the data and tests significance
by permuting ancestry profiles for the overall data set and for pairwise
combinations of sampled populations. Finally, ObStruct produces three CDA
plots which visualize the clustering of individuals by sampled population
within your data in a manner similar to a PCA plot.
ObStruct is implemented in Perl and is available from our lab website at: