The analysis of large data sets describing reproductive isolation between species that vary in their degrees of relatedness has been extremely influential in the study of speciation. However, several limitations make it difficult to test specific hypotheses about which factors predict the evolution of reproductive isolation. In particular, the statistical methods typically used are limited in their ability to test complex hypotheses involving multiple predictor variables or interactions between variables; at least one method, the Mantel Test, has also been found to be unreliable. In this paper, I describe a framework to determine which factors contribute to the evolution of reproductive isolation using phylogenetic linear mixed models. Phylogenetic linear mixed models do not suffer from the same statistical limitations as other methods and I demonstrate the flexibility of this framework to analyze data collected at different evolutionary scales, to test both categorical and continuous predictor variables, and to test the effect of multiple predictors simultaneously, all of which cannot be achieved using any other single statistical method. I do so by re-analyzing several classic data sets and explicitly testing hypotheses that had previously been untested directly, including differences in accumulation of reproductive isolation between sympatric and allopatric species pairs.