FPtest

Fligner-Policello non-parametric two-samples robust-rank order test.

Fligner-Policello is a non-parametric test of two combined random variables with continuous cumulative distribution. It is a robust rank-order test of treatment effects for populations which assumes neither normality nor equal variances (Behrens-Fisher problem). It does one assumption, which is satisfied when the distributions are symmetric.

File only needs to input the data matrix (n-by-2: data=column 1, samples=column 2), specified direction and alpha significance level.

It outputs a complete table with the Mann-Whitney-Wilcoxon, Fligner-Policello and Fligner-Policello variance statistics, and the cumulative probability value associated to the Fligner-Policello statistic.

For the associated P-value there are two main cases: 1) if (nx <= 12) & (ny <= 12); it calls to a two possible Mann-Whitney's U cumulative distribution functions, asking for an exact value (a bit long running time) or approximate value (immediate procedure); 2) if (nx > 12) & (ny > 12), procedure is based on the standard normal distribution.