Caracterização da análise estatística das publicações da área de ciências do esporte em um intervalo de 20 anos: um estudo retrospectivo descritivo e comparativo

Abstract

General objective - to identify and compare the statistical tests used in scientific articles published in journals in the field of Sport Sciences over a 20-year interval, assessing whether the statistics used in the articles have been updated. Specific objectives - to identify statistical tests used in publications from the 2000s and 2020s and the changes that occurred; identify general data from journals: country of author's home institution, number of articles published per year and impact factor; identify general data from the articles and the sample. Methods - a retrospective, descriptive, exploratory study of bibliographic research with a quantitativequalitative approach. Selection of journals – based on the highest impact factor value (JCR, Journal of Citation Reports) 15 scientific journals in the field of Sport Sciences that published research on the theme “human performance” and that met the following criteria were included in the study: ) Sports Science area; b) studies evaluating physical performance; c) access to the collection of articles through the databases available for research or free access; d) publication of original articles in 2020 and in a time window of approximately 20 previous years. Exclusion criteria: a) start of publication after 2001; b) narrative review, systematic review, meta-analysis, editorials, opinions, animal model, disease or research that do not focus on the theme “physical performance”. Selection of articles – up to a maximum number of 50 articles per journal in the years 2000 and 2020. Inclusion criteria: a) original articles published in 2000 and 2020 by the same journal; b) related to the theme “physical activity”; c) presentation of the description of the statistical methods used in the study. Exclusion criteria: a) narrative review, systematic review, meta-analysis, editorials and opinions; b) with a predefined statistical proposal or that did not present it; c) studies with animals and/or cadavers; d) without full access. Data extracted from articles - edition, year and title; nationality of the corresponding author's institution; sample references (quantitative, sex, average age, age group and number of groups); sample characterization (participants' fitness level and type of physical modality practiced); intervention that generated the research result (type of main intervention performed and intervention moments); information on the general statistical methodology used (sample calculation, effect size, significance level, normality test and descriptive statistics - mean, standard deviation and percentage); reported statistical software and tests (grouped by specificity and individualized by name). Conclusion – We analyzed 1,205 articles from 15 scientific journals from the 2000s and 2020s, seeking to understand whether the statistics used to describe the results of the studies changed in the period. We conclude that there were no changes in the use of statistical tests. The tests used in 2020 already existed in 2000, and no new procedures were found. Comparison, correlation, association, and post hoc tests were well used in both years. In 2020, the use of normality, homogeneity, regression and effect size tests was expanded, which were not frequent in 2000. Changes in the use of sample calculation, effect size, significance level, descriptive statistics and normality test were also were found, and these items were few documented in 2000. It is possible to notice that the statistical methods of the 2020 articles were more complex and produced more elaborate reports in the validation of the results.