Reporting transparency in analysis of variance assumption testing

ANOVA models in general are valid only insofar as both homoscedasticity and multiple comparisons between groups are controlled for. While recent studies suggest that researchers in various disciplines seldom comply with these requirements, lack of rigor in methodological reporting makes a proper, large-scale, and transdisciplinary assessment of the situation impossible. Given this situation, the present paper attempts a large-scale and transdisciplinary assessment of statistical reporting transparency in ANOVA-related papers, using customized regular expressions on the full-text of articles published in Elsevier journals. Results show that, beyond important variations at multiple disciplinary levels, reporting transparency in the testing of ANOVA statistical assumptions is generally lacking, as a minority of ANOVA-related papers refer to either homoscedasticity or multiplicity correction. Also, articles from higher-quartile journals tend to report statistical assumptions more thoroughly than those from lower-quartile journals, regardless of discipline or subdiscipline.