Sekce 1 – Teorie a metoda v pedagogice
Autoři: Petr Soukup, Ilona Kočvarová
Abstrakt: Cílem příspěvku je poukázat na skutečnost, že při použití chi-kvadrát testu nezávislosti pro běžně velké datové soubory je síla použitého testu velice malá a na výsledky tak nelze příliš spoléhat.
Klíčová slova: Kontingenční tabulka, síla testu, chi-kvadrát test
Title: How big sample size is necessary for categorical data analysis? (About „weakness“ of chi-square test)
Abstract: Contingency tables and chi-square test of independence are among the most popular tools of researchers in educational research. The question is whether it is really a universal tool suitable for analyzing social science data. Apart from the situations in which tests are used for inappropriate data in terms of obtaining a sample (particularly data from convenience samples), it is appropriate to analyze sample size requirements for correct application of the chi-square test of independence. Based on the power of the test (Cohen 1988, Kraemer & Blasey, 2016), it will be shown that for common sample sizes the chi-square test is „weak“ (showing low test power) and is therefore not a suitable tool for detecting context. Computing illustrations will come from IBM SPSS Sample Power, and G * Power will provide comparable results. At the end of the paper, alternatives for testing categorical data will be discussed.
Keywords: Contingency table, power, chi-square test
Download the article in English [PDF]: Not available