A graphical method of cumulative differences between two subpopulations

Abstract Comparing the differences in outcomes (that is, “dependent variables”) between two subpopulations is often most informative when comparing only for individuals from who are similar according to “independent variables.” The independent variables generally known as “scores,” propensity scores matching or probabilities predicted by statistical machine-learned models, example. If discrete, then some averaging necessary reduce noise arising varying randomly over those discrete values observed data. traditional method of bin data and plot average outcome each against score bin. However, such binning can be rather arbitrary yet greatly impacts interpretation displayed deviation assessment its significance. Fortunately, entirely unnecessary plots cumulative associated scalar summary metrics that analogous workhorse statistics probability distributions—those due Kolmogorov Smirnov their refinements Kuiper. present paper develops methods common case which no any member being compared exactly equal other either subpopulation.