Аннотация:Testing sample distinction is necessary in a wide range of practical tasks. Medicine, sociology, psychology, marketing - this is a short list of industries where it is required to conduct tests that establish effectiveness or inefficiency of a certain technology.
Diversity of situations and techniques applied to sample distinction create a problem for compliance of testing procedures. The problem rises for tests including large and small samples (dependent or independent) with various distributions.
The article proposes a classification of problems created by testing differences between two samples. Limits of applicability of parametric and non-parametric tests are established based on selected distribution. The classification is presented in table of correspondences between the problems encountered and corresponding testing technique. Informative examples are included based on simulated and real data.
SPSS software was used for sample distinction tests. SPSS algorithms are not mathematically justified. It is important to double-check the operation of the machine computing procedure "manually” to understand the nature of tests, area of applicability and determine parameter limits. The article provides mathematical justification for the algorithms used, which can be considered as supplementary information missing in SPSS software.