Researcher Degrees of Freedom and a Lack of Transparency Contribute to Unreliable Results of Nonparametric Statistical Analyses Across SPSS, SAS, Stata, and R
The replication crisis within the social and behavioral sciences has called into question the consistency of research methodology. A lack of attention to minor details in replication studies may limit researchers’ abilities to reproduce the results. One such overlooked detail is the statistical programs used to analyze the data. In the current investigation, we compared the results of several nonparametric analyses and measures of normality conducted on a large sample of data in SPSS, SAS, Stata, and R with results obtained through hand-calculation using the raw computational formulas. Multiple inconsistencies were found in the results produced between statistical packages due to algorithmic variation, computational error, and lack of clarity and/or specificity in the statistical output generated. We also highlight similar inconsistencies in supplementary analyses conducted on subsets of the data, which reflect realistic sample sizes. These inconsistencies were largely due to algorithmic variations used within packages when the analyses are performed on data from small- or medium-sized samples. We discuss how such inconsistencies may influence the conclusions drawn from the results of statistical analyses depending on the statistical software used, and we urge researchers to analyze their data across multiple packages, report details regarding the statistical procedure used for data analysis and consider these details when conducting direct replications studies.