Note on the applicability of the F-model in analysis of pollen pool heterogeneity

Measuring genetic structure of pollen gametes has become the standard approach in studies on plant mating system and pollen flow. However, the popular statistical procedure, TwoGener, provides the point estimate of the pollen pool heterogeneity based on AMOVA, implying any statistical tests regarding pollen pool structure to be based on permutation procedures. In this article, the Bayesian approach was developed based on the Dirichlet-multinomial model (the so-called F-model). Computer simulations revealed the applicability of the Bayesian method in the analysis of pollen pool heterogeneity. Compared with TwoGener, the Bayesian method appeared to have better statistical properties when estimating the inverse of heterogeneity index, mostly due to the relatively high accuracy of the Bayesian estimate when the actual heterogeneity is low. Hence, the Bayesian approach may be preferred, for example, when the effective number of pollen parents is to be assessed. Other applications and potential further developments of the Bayesian method are also briefly discussed.