Frequently occurring errors in agricultural research due to the incorrect use of statistical methods

Abstract

The objectives of this article were I) to warn about recurrent statistical errors that are frequently found in different scientific papers in agricultural research and associated with conceptual aspects and the misuse of computational tools, specifically, in the context of the analysis of variance (ANOVA) and II) provide work methodologies that allow the correct use of these statistical procedures. Four types of errors are approached: an error associated with the assumption of normality; a second one linked to non-parametric multiple comparison tests; another error consists of not considering a single standard error when reporting confidence intervals in an ANOVA context; and the fourth error is related to the interpretation of the p-value when it is very close to the level of significance. An example is included and shows how to proceed by using statistical software. In general, criteria and work methodologies are established that favor the correct use of these procedures. Taking into consideration the aspects approached in this work will make it possible to avoid issuing biased conclusions, which will contribute to obtaining significant and valuable information from the results in research processes and scientific papers.