Calculating and reporting effect sizes on scientific papers (1): p < 0.05 limitations in the analysis of mean differences of two groups

• Autor(a) principal: Espirito Santo, Helena
• Data de Publicação: 2015
• Outros Autores: Daniel, Fernanda Bento
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://doi.org/10.7342/ismt.rpics.2015.1.1.14
• topic: Tamanho do efeito; Significância estatística; Valor p; d de Cohen; g de Hedges; Delta de Glass; Effect size; Statistical significance; Cohen's d; p-value; Hedges’ g; Glass’s Delta

The Portuguese Journal of Behavioral and Social Research requires authors to follow the recommendations of the Publication Manual of the American Psychological Association (APA, 2010) in the presentation of statistical information. One of the APA recommendations is that effect sizes should be presented along with levels of statistical significance.&nbsp;Since p-values from the results of the statistical tests do not indicate the magnitude or importance of a difference, then effect sizes (ES) should be reported. In fact, ES gives meaning to statistical tests; emphasizes the power of statistical tests; reduces the risk of interpreting mere sampling variation as real relationship; can increase the reporting of “non-significant" results, and allows the accumulation of knowledge from several studies using meta-analysis.&nbsp;Thus, the objectives of this paper are to present the limits of the significance level; describe the foundations of presentation of ES of statistical tests to analyze differences between two groups; present the formulas to calculate directly ES, providing examples of our own previous studies; show how to calculate confidence intervals; provide the conversion formulas for the review of the literature; indicate how to interpret the ES; and show that, although interpretable, the meaning (small, medium or large effect for an arbitrary metric) could be inaccurate, requiring that interpretation should be made in the context of the research area and in the context of real-world variables.