A0688
Title: A goodness-of-fit test for the geometric maximum compound logistic distribution model
Authors: Ludwig Baringhaus - Leibniz University Hannover (Germany)
Daniel Gaigall - FH Aachen University of Applied Sciences (Germany) [presenting]
Abstract: Based on independent copies of a bivariate random vector $(M,N)$, with positive integer-valued component $N$, testing the composite hypothesis that $(M,N)$ follows a geometric maximum compound logistic distribution model. This distributional model is of interest, for example, in hydrology, where $N$ models the number of floods and $M$ is the maximum flood water level during a certain time period. The geometric maximum compound logistic distribution model is characterized in the sense that a special transform of $(M,N)$ fulfils a specific equation. A weighted integral of an expression is suggested, obtained by replacing the function part of this equation with empirical counterparts as test statistics and proposing a parametric bootstrap procedure to get critical values. A simulation study shows the performance of the new procedure. The test is applied to a hydrological data set. A new goodness-of-fit test for the logistic distribution is obtained as a special case of the novel approach.
