Traditionally, the underlying assumptions made in using r as a measure of association are that (1) the individuals in the sample are statistically independent of each other, and (2) the population from which the sample was drawn has a vi variate normal distribution for the two traits of interest (Whitlock & Schluter,2009).