Kernel Testing as an Alternative to χ Analysis for Investigating the Distribution of Quantitative Traits

Chi-square analysis is a popular means of analyzing data from inheritance studies. Such an analysis, though, can be inappropriate when the phenotypic characteristic under study is a continuous variable. We argue that arbitrary categorization of continuous variables can either seriously invalidate or make powerless a chi-square goodness-of-fit test. We propose an alternative test based on kernel density estimates. The test makes no assumption about the underlying distribution of the phenotypic variables. Furthermore, kernel density estimates provide an appealing means of describing the distribution of continuous variables.