Our preference here is for greater geographic
coverage and more sampled populations than for high precision in
the estimates of the relationships, as the main source of uncertainty
for our analytical approach is going to be due to the number of
groups used to fit the model and the mismatches among groups
(see below).