NPST: A generalized non-parametric seasonality test

Roland Rau, Max Planck Institute for Demographic Research

This paper represents a generalized version of the previous nonparametric seasonality tests by Hewitt et al. (1971) and Rogerson (1996) which relaxes their respective assumptions. An algorithm has been developed which allows to empirically determine the distribution of the test statistic using Monte Carlo simulation. This algorithm has been implemented in various computer languages and tentatively been called NPST which stands for Non-Parametric Seasonality Test. The implementations have been checked for correctness and benchmarked for speed of executing the code. Tests have been conducted to determine how many Monte-Carlo samples have to be generated to approximate the true distribution of the test statistic. The respective software can be obtained from the author.

  See paper

Presented in Session 73: Various methods