Data driven chi-square test for uniformity with unequal cells. (English) Zbl 1054.62057

Summary: The power of Pearson’s chi-square test for uniformity depends heavily on the choice of the partition of the unit interval involved in the form of the test statistic. We propose a selection rule which chooses a proper partition based on the data. This selection rule leads usually to essentially unequal cells well suited to the observed distribution. We investigate the corresponding data driven chi-square test and present a Monte Carlo simulation study. The conclusion is that this test achieves a high and very stable power for a large class of alternatives, and is much more stable than any other test we compare to.


62G10 Nonparametric hypothesis testing
65C05 Monte Carlo methods
62G20 Asymptotic properties of nonparametric inference
Full Text: DOI