Data transformations for ANOVA assumption violations in agrarian sciences subfields: a systematic review, simulation study, and practical guidelines

The use of data transformations for validating variance analysis has been a technique adopted by the scientific community since the 20th century. In this study, we conducted a systematic review covering four subareas of agricultural sciences (entomology, plant science, forestry, and soil science) to investigate how this technique has been applied in these fields. In addition to the systematic review, we conducted a simulation study to evaluate the effectiveness of logarithmic, square root, arcsine, and Box-Cox transformations in correcting violations of the assumptions of normality and homogeneity of variances, as these were the most frequently used transformations in the systematic review. Although some transformations effectively homogenize variances, few correct both non-normality and heterogeneity simultaneously. To facilitate the screening phase in the systematic review, we developed a web application as an alternative for users who are not familiar with or uncomfortable using Python's interface. Finally, we offer practical guidelines for applying ANOVA with transformations, providing the audience with a helpful tool to minimize the errors and misconceptions identified in the systematic review.