Research on K Distribution in Anderson-Darling Goodness-of-Fit Test

Abstract

Identification of radar clutter model is one important step to choose optimum CFAR processor in radar signal detection, and Anderson-Darling (AD) goodness-of-fit test considered the practical and effective way under small sample size. As K distribution is a complex model providing suitable description for the amplitude of microwave sea clutter, the research on K distribution in AD test is useful but very difficult. Moment-based methods are widely used to estimate parameters of K distribution and comparison show that methods with lower moments are more accurate. Further in AD test, the influence of shape parameter on critical value (CV) for K distribution is presented. Taking it into account, tables of AD test results using varied critical value advanced in our study and fixed critical value proposed in previous research are shown based on Monte Carlo simulations.