SIMULATION STUDIES OF THE KOLMOGOROV-SMIRNOV TEST ON DISTRIBUTIONS WITH UNKNOWN PARAMETERS.

Abstract : The standard Kolmogorov-Smirnov (K-S) test is valid when the null distribution is completely specified. Lilliefors (1967) shows for the case when the null distribution is normal, but the mean and variance are unknown and must be estimated from the data, that the K-S test is extremely conservative. This research constructs by Monte Carlo simulation tables of critical values for the K-S statistic when the null distribution is either negative exponential or rectangular with unknown parameters that are to be estimated from the data. The results indicate also that the standard K-S test is conservative. Moreover, power studies are simulated in both cases. These studies indicate the danger of using the K-S test for small samples. (Author)