Calcular e apresentar tamanhos do efeito em trabalhos científicos (2): Guia para reportar a força das relações.

Abstract

No primeiro número da Revista Portuguesa de Investigação Comportamental e Social foi descrita a importância de calcular, indicar e interpretar os tamanhos do efeito para as diferenças de médias de dois grupos. A RPICS pretende continuar a alertar para a importância de reportar os tamanhos do efeito para outros testes estatísticos. A magnitude da força das relações não foi indicada no artigo prévio e talvez não se saiba que a correlação é um tamanho do efeito. Assim, este artigo pretende fornecer algumas diretrizes aos autores sobre os procedimentos de cálculo do coeficiente de correlação de Pearson e alguns coeficientes de correlação para dados especiais (correlação de Spearman, Tau de Kendall, Coeficiente de correlação ponto-bisserial e bisserial, Coeficiente Fi, V de Cramér e Eta). Com esse objetivo, serão apresentadas as fórmulas, os passos no SPSS (Statistical Package for the Social Sciences), pressupostos e precauções, classificação dos valores e sua interpretação. Uma vez que o SPSS não computa todos os testes referidos, nos suplementos ao artigo são incluídas cinco folhas de cálculo (3 formas de comparar correlações, correlações ponto-bisserial e bisserial e correção de correlações para amostras < 60). || In the first issue of the Portuguese Journal of Behavioral and Social Research, it was described the importance of calculating, indicating and interpreting the effect sizes for the differences between means of two groups. The PJBSR intends to continue to remind of the importance of reporting effect sizes for other statistical tests. The magnitude of the strength of the relationships was not indicated in the previous paper, and it may not be known that correlation is an effect size. Thus, this paper intends to provide some guidelines to the authors on the procedures for calculating the Pearson correlation coefficient and some correlation coefficients for special data (Spearman correlation, Tau de Kendall, Point-biserial and biserial correlation coefficient, Fi coefficient, Cramér, and Eta). For this purpose, the formulas, steps in the SPSS (Statistical Package for the Social Sciences), assumptions and precautions, classification of values and their interpretation will be presented. Since SPSS does not compute all the mentioned tests, five spreadsheets (3 ways of comparing correlations, point-biserial and biserial correlations, and correction of correlations for samples < 60) were included in the article supplements.

In the first issue of the Portuguese Journal of Behavioral and Social Research, it was described the importance of calculating, indicating and interpreting the effect sizes for the differences between means of two groups. The PJBSR intends to continue to remind of the importance of reporting effect sizes for other statistical tests. The magnitude of the strength of the relationships was not indicated in the previous paper, and it may not be known that correlation is an effect size. Thus, this paper intends to provide some guidelines to the authors on the procedures for calculating the Pearson correlation coefficient and some correlation coefficients for special data (Spearman Rho, Kendall’s Tau, Point-biserial and biserial, Phi, Cramér’s V, and Eta). For this purpose, the formulas, steps in the SPSS (Statistical Package for the Social Sciences), assumptions and precautions, classification of values and their interpretation will be presented. Since SPSS does not compute all the mentioned coefficients, five spreadsheets (3 ways of comparing correlations, point-biserial and biserial, and correction of correlations for samples < 60) were included in the article supplements.