Proportion reduction in error (PRE) is the proportion of error that has been reduced by a more complex model compared with a simpler model, which in our course is always the empty model. When comparing to the empty model, PRE is calculated as SS Model (the sum of squares reduced by the model) divided by SS Total (the total sum of squares in the outcome variable under the empty model). When the comparison is with the empty model, PRE is the same as Eta-squared (in ANOVA) and R^2 (in regression).