Practical Predictive Analytics

Point biserial correlation

Another basic technique that you could use if one of the categories is a nominal variable with only two classes and the other variable is quantitative would be point biserial correlation. However, since this technique uses Pearsons correlation coefficient, you need to make certain assumptions about the distributions of the data. For example, the data needs to be normally distributed and have equal variance. This assumption holds for our simulated sales example, since each category is generated using the rnorm() function which simulates 2 randomly distributed data frames with different means, but identical standard deviations (variances)

To show point biserial correlation, we can use our previous sales treatment. ...