Infer\^encia Baseada em Magnitudes na investiga\c{c}\~ao em Ci\^encias do Esporte. A necessidade de romper com os testes de hip\'otese nula e os valores de p

TL;DR

This paper critiques the reliance on null hypothesis significance testing in sports science research and advocates for magnitude-based inference to better interpret practical effects.

Contribution

It proposes replacing p-value-based significance tests with magnitude-based inference methods for more meaningful interpretation of results in sports science.

Findings

Highlights limitations of p-value significance testing

Advocates for magnitude-based inference for practical interpretation

Discusses limitations and benefits of magnitude-based approaches

Abstract

Research in Sports Sciences is supported often by inferences based on the declaration of the value of the statistic statistically significant or nonsignificant on the bases of a P value derived from a null-hypothesis test. Taking into account that studies are manly conducted in sample, the use of null hypothesis testing only allows estimating the true values (population) of the statistics used. However, evidence has grown in many areas of knowledge that this approach often leads to confusion and misinterpretation. To overcome this limitation they have recently emerged recommendations to support the statistical analysis with approaches that make use of more intuitive interpretations and more practical, especially based on the magnitudes (certainty / uncertainty) of the true values found. With the intent to provide alternative solutions to methodological designs recurrently used in research in sports sciences, this paper will seek to i) briefly spell out some of the weaknesses associated with the null hypothesis tests based in the P value; ii) reflect on the implications of the use of practical/clinical significance as opposed to statistical significance; iii) submit proposals for use the inferences based on the magnitude, particularly in the visualization and interpretation of results; iv) present and discuss the limitations of magnitude-based inference. Thus, this update article discourages, in a sustained-based, the use of significance tests based only on the concept of null hypothesis. Alternatively, it is proposed to use methods of inference based on magnitudes as they allow interpretations of the practical/clinical effects results obtained. ----- As investiga\c{c}\~oes em ci\^encias do esporte sustentam-se frequentemente em infer\^encias baseadas na declara\c{c}\~ao de um valor estatisticamente significativo, ou n\~ao significativo, com base no valor de p que deriva dos testes de hip\'otese nula. Considerando que os estudos s\~ao iminentemente amostrais, o recurso aos testes de hip\'otese nula apenas possibilita estimar os valores verdadeiros (popula\c{c}\~ao) das estat\'isticas utilizadas. Contudo, tem crescido a evid\^encia, em diversas \'areas do conhecimento, de que esta abordagem origina frequentemente interpreta\c{c}\~oes confusas e at\'e erradas (7). Para ultrapassar esta limita\c{c}\~ao t\^em surgido recentemente recomenda\c{c}\~oes no sentido de sustentar as an\'alises estat\'isticas com abordagens que recorram a interpreta\c{c}\~oes mais intuitivas e mais pr\'aticas, baseadas sobretudo nas magnitudes (certezas/incertezas) dos valores verdadeiros encontrados. Com o intento de fornecer pistas alternativas aos desenhos metodol\'ogicos recorrentemente utilizados na investiga\c{c}\~ao em ci\^encias do esporte, neste trabalho procuraremos i) enunciar sucintamente algumas das fragilidades associadas aos testes de hip\'otese nula sustentados no valor de p; ii) refletir sobre as implica\c{c}\~oes da utiliza\c{c}\~ao da signific\^ancia pr\'atica/clinica em oposi\c{c}\~ao \`a signific\^ancia estat\'istica; iii) apresentar propostas de utiliza\c{c}\~ao das t\'ecnicas de infer\^encias baseadas na magnitude, particularmente na visualiza\c{c}\~ao e interpreta\c{c}\~ao dos resultados; iv) apresentar as principais limita\c{c}\~oes do uso das infer\^encias baseadas em magnitudes. Assim, neste artigo de atualiza\c{c}\~ao desencoraja-se, de forma sustentada e fundamentada, o uso dos testes de signific\^ancia baseados apenas no conceito de hip\'otese nula. Em alternativa, prop\~oe-se a utiliza\c{c}\~ao de m\'etodos de infer\^encias baseados em magnitudes por possibilitarem interpreta\c{c}\~oes dos efeitos pr\'aticos/cl\'inicos dos resultados obtidos.