A review of the statistical methods used in sports and exercise sciences PhD theses: a case study in a single post-graduate program

Abstract

This study aimed to describe the statistical methods used and statistical practice in PhD thesis in sports and exercise sciences, and to examine the time trends of statistical methods prevalence. We analyzed 50 PhD theses supported by a post-graduate program recognized by CAPES and defended between January 2003 and December 2013. Most theses used more than two statistical methods, and T-test, ANOVA, multiple comparisons, correlation and reliability analysis were the most used approaches. Less than 50% of both experimental and observational theses reported the effect size (ES), confidence interval (CI), sample size calculation and observed power analysis. The use of multiple comparisons in experimental studies significantly increased in the last 5-years compared to 2004-2008 time period. No other significant changes were observed. Results suggest that several statistical methods are used in quantitative design studies. Due to the observed large prevalence of statistical deficiencies (lack of reporting ES, CI, power analysis), postgraduate programs should have qualified statisticians with a major say in checking the statistical quality of PhD theses and subsequent articles.