Valerii Bosniuk
Ph.D. in Psychology, Associate Professor. National university of civil defence of Ukraine
ORCID 0000-0003-0141-1920
DOI https://doi.org/10.52363/dcpp-2021.2.9
Keywords: effect size, effect size estimation, null hypothesis significance testing, p-value
The procedure for testing the significance of the null hypothesis (common abbreviation NHST - Null Hypothesis Significance Testing) is have being used for confirmation of research results in psychological research for many years and it also includes special statistical criteria. At the same time, for the most part, the value of Statistics «p» (p-value) is considered as the equivalent of the importance of the results obtained and the strength of scientific evidence in favor of the practical and theoretical effect of the study. This incorrect use and interpretation of p-value impugns the application of Statistics in general and threatens the development of psychology as a science. The identification of a statistical conclusion with a scientific conclusion, the focus solely on novelty in scientific research, the ritual commitment of researchers to the significance level of 0.05, the reliance on the statistical categorical "yes/no" when making a decision leads to the fact that psychology multiplies only the results about the presence of an effect without taking into account its magnitude, practical value.
This work is intended to analyze the limitations of p-value when interpreting the results of psychological research and the benefits of presenting information about the effect size. Applying the size of the effect will allow you to make the transition from dichotomous thinking to evaluative thinking, determine the value of results regardless of the level of statistical significance, and make decisions more rationally and reasonably.
The position that the author of the scientific work in formulating the study conclusions should not be limited to a single indicator of the level of statistical significance is substantiated. Reasonable conclusions should be based on a reasonable balancing of p-value and other equally important parameters, one of which is the size effect. The effect (difference, connection, association) can be statistically significant, and its practical (clinical) value – insignificant, trivial. «Statistically significant» does not mean «useful», «important», «valuable», «considerable». Therefore, the attention of psychologists to the analysis of the detected size effect should be mandatory when interpreting the results of the study.