A Nonparametric Approach to Program Evaluation: Utilizing Number Needed to Treat, L’Abbé Plots, and Event Rate Curves for Outcome Analysis

Abstract

The medical sciences offer several helpful tools to the program evaluator for outcome analysis. Where a single point on the range of a dependent variable clearly defines attainment of outcome success, the number needed to treat (NNT) offers an indicator of a program’s efficacy. L’Abbé plots compare several programs in terms of the event rates that constitute the NNTs. If the nature of an outcome variable is not amenable to such a demarcation, then the evaluator may extend the NNT concept by calculating and plotting event rates on a continuous, ordinal basis. Graphical analysis of these curves may involve program comparisons, inclusion of other relevant variables, and visualization of the test statistic for inferential purposes.