Referential models have also been extensively used to argue that some traits are more useful for reconstructing population and species relationships (Smith et al., 2007, 2013; von CramonTaubadel and Smith, 2012).Our results suggest that none of the traits examined herein are more reflective of population structure than others because the posterior credibility intervals of the estimates
of the proportion of variance attributable to population structure overlap quite a bit. Rogers and Harpending (1983) pointed out that a single completely heritable characteristic has as much evolutionary information as a single dinucleotide locus, which is to say, very little. The conclusion that we draw from the combination of this theoretical insight with our empirical result is that there is no good reason to believe that we will be able to identify characteristics that are better for reconstructing phylogeny or population history. Even if some traits were more useful in a given set of populations, there is no theoretical reason to suppose that this is a general property across all groups we study. In light of this, we conclude that limiting characteristics for use in phylogenetic or population structural analyses based on criteria from comparative studies likely excludes perfectly good data.