A statistical mean dyad test (tt-criteria for comparison of mean trends of two samples)
Genkin, A.A.; Emanuél', V.L.
Klinicheskaia Laboratornaia Diagnostika 3: 45-47
1994
ISSN/ISBN: 0869-2084 PMID: 7921822 Document Number: 432675
The authors claim that widely used statistic tests (Student, Berence-Fisher, Wilcoxon-Mann-Whitney, etc.) may fail to detect differences in average trends. A new statistical tt-test is suggested for such situations when samples are compared belonging to essentially different distributions and one of these distributions is bimodal or U-shaped. This test consists in comparison of the arithmetic means calculated not for the whole sample but separately for values higher and lower than the respective medians. The authors persuasively illustrate the efficacy of this method used in analysis of routine clinico-laboratory data. The statistical mean dyad test (tt-test) is realized in GEMA intellectual system statistical module and used automatically only in cases when routine methods are ineffective.