computing the two-sided kolmogorov-smirnov distribution.pdfVIP

JSS Journal of Statistical Software MMMMMM YYYY, Volume VV, Issue II. / Computing the Two-Sided Kolmogorov-Smirnov Distribution Richard Simard Pierre L’Ecuyer Universit´e de Montr´eal Universit´e de Montr´eal Abstract We propose an algorithm to compute the cumulative distribution function of the two- sided Kolmogorov-Smirnov test statistic Dn and its complementary distribution in a fast and reliable way. Different approximations are used in different regions of (n, x). Java and C programs are available. Keywords : KS test, Kolmogorov-Smirnov distribution, goodness-of-fit. 1. Introduction The Kolmogorov-Smirnov (KS) two-sided test statistic Dn is widely used to measure the goodness-of-fit between the empirical distribution of a set of n observations and a given continuous probability distribution. It is defined by ˆ D = sup G(x) − G (x) , n n x ˆ where n is the number of (independent) observations, Gn is their empirical cumulative dis- tribution function (cdf), and G is a completely specified continuous theoretical cdf. Let Fn denote the cdf of Dn under the null hypothesis H0 that the n observations are independent and have cdf G, that is, F (x) = P[D x | H ] for