Coalescent approaches to phylogeography and approximate Bayesian Computation

Last time we saw an early example of using coalescent theory to distinguish between two scenarios describing the history of populations. In the example we considered, Knowles [9] compared two scenarios, the “widespread ancestor” and the “multiple glacial refugia” scenarios. To make the comparison she simulated data under the “widespread ancestor” hypothesis, collected the samples into a multiple-refuge tree, and calculated a statistic that measures the discrepancy between the gene trees and the population trees. Her observed gene tree was far less discordant than the simulated trees, leading her to conclude that her grasshoppers had been dispersed among multiple refugia in the past rather than being the remnants of a single, widespread ancestral population. As I mentioned, one limitation of the approach Knowles [9] takes is that it requires the investigator to identify alternative scenarios before beginning the analysis, and it can only identify which of the scenarios is more likely than the others with which it is compared. It cannot determine whether there are other scenarios that are even more likely. Another approach is to back off a bit, specify a particular process that we are interested in and to use what we know about that process to try and estimate its properties.