Kolmogorov-Smirnov test for grouped data

Abstract

Herein is presented a Kolmogorov-Smirnov test for grouped samples. It constitutes an application of W. J. Conover's procedure, which was originally designed for calculating exact critical levels for test with discrete distributions. This test extends the goodness-of-fit test to samples which are grouped into intervals. Critical levels in the two-sided case were calculated to a close approximation. Some examples of the application of this extension of the Kolmogorov-Smirnov test are included and comparisons with the Chi-square test, the Kolmogorov-Smirnov test for continuous distribution and for grouped samples are made. Comparisons among the three goodness-of-fit tests based on the results of computer simulations are made.