An Application of a Variable Window Discrete Scan Statistic in Molecular Biology

摘要

In this article a variable window discrete scan statistic is introduced for a conditional scan statistic. This statistic is used to test the null hypothesis that the observations are randomly distributed according to an i.i.d. Bernoulli model, the total number of successes is known, against an alternative hypothesis that suggests clustering in a sequence of unknown size. Approximations for critical values for specified significance levels are evaluated based on product-type approximation and Monte Carlo methods. An example from an experiment in molecular biology, where a sequence of amino acids is examined for an unusual large negative or positive charge is presented. We evaluate the performance of this variable window discrete scan statistic by comparing its power with various fixed length window scan statistics.