Gene Conservation Laboratory
Statistics Program for Analyzing Mixtures (SPAM) Software


SPAM software logo

A Comment on Parameter Identifiability

Consider two regions, each w/ two populations. Assume the populations w/in a region are genetically similar, but the regions are genetically distinct. While one can clearly identify what contribution Region 1 makes to a mixture, it may be less clear what the break down of contributions is for the populations making up Region 1. This is a problem of parameter identifiability - we can get good estimates of the sum of population contributions, but not of the individual population contributions.

This is akin to the problem of multicollinearity in multiple regression, and can generally be detected by very high bootstrap variation of the individual parameter estimates compared to the bootstrap variation of the region estimates. Another flag would be having the mean bootstrap estimate for each of the regions be near its expected value, but the mean bootstrap estimates for the population contributions diverging from their expected values.