Directional Tests for Ordered Categorical Data: a New Permutation Solution.
by
Stefano Bonnini
Department of Statistics - University of Padova
Coauthors: Rosa Arboretti Giancristofaro Luigi Salmaso

Frequently in the modern socio-economic systems the aim of a performance analysis or a quality evaluation is to compare different products, different manifacturing plants or service centres, different actions and comparing two of them we need to answer the question “which is better?”. This evaluation is complicated because the aspects to be considered are often several, different and measured through categorical data. From the statistical point of view it is a hypotheses testing problem.

We discuss some directional permutation tests (based on the nonparametric combination of dipendent permutation tests) for the problem of two sample comparison in presence of ordinal categorical variables.

In particular we present a new permutation test based on the combination of a finite number of sample moments.

We also show the results of Montecarlo simulations to compare permutation tests with other nonparametric tests.

Date received: July 11, 2006