NEW POLYSTOCHASTIC STATISTICAL INFERENCE IN SOCIAL SCIENCES - DEFINING NEW RULES AND THRESHOLDS

Abstract

The Null Hypothesis Significance Testing (NHST) framework has sparked considerable debate within the scientific community, leading to numerous studies advocating for a re-evaluation of the current system. New polystochastic statistical inference defines methods of statistical inference that integrate rules and thresholds for both rejecting the null hypothesis and confirming the alternative hypothesis. This approach unifies the control of respondents' influence on statistical significance and introduces criteria such as effect size and Bayesian inference for confirming the alternative hypothesis. Unlike NHST, polystochastic statistical inference controls Type I error (p-value) and aims to optimize the confirmation of evidence without increasing the risk of Type II errors.

Alternate abstract:

Okvir testiranja pomembnosti nicelne hipoteze (angl. Null Hypothesis Significance Testing - NHST) je sprozil precejánjo razpravo у znanstveni skupnosti. To je vodilo do stevilnih studij, ki zagovarjajo ponovno oceno sedanjega sistema. Novo polistohastièno statisticno sklepanje definira metode statistienega sklepanja, ki zdruZujejo pravila in pragove tako za zavracanje nicelne hipoteze kot za potrditev alternativne hipoteze. Ta pristop poenoti nadzor nad vplivom anketirancev na statistiéno pomembnost in uvede merila, kot sta velikost ucinka in Bayesov sklep za potrditev alternativne hipoteze. Za razliko od NHST polistohastiéno statisticno sklepanje nadzoruje napako tipa I (p-vrednost) in Zeli optimizirati potrditev dokazov brez povecanja tveganja napak tipa II.