What statistic to use from pplacer to decide on the best reference set among several reference sets?

[Ramzi Alsallaq]

Dear pplacer developers and community users,

I am using several reference data sets that are subsets of each other (say set1 includes set2 and set2 includes set3 and so on: set1>set2>set3>..) to place the same queries using pplacer and the same exact pipeline for all three to minimize any variations other than the differences between these sets. I am thinking of this problem as having hierarchically nested models and my question is what statistic(s) from pplacer can be used to decide on the best model.  Any ideas would be appreciated.