Improving Research through Avoiding Common Statistical Errors: The Case of Piosphere

Abstract

For many years scientists studied the piosphere concept- a grazing gradient around a natural/artificial watering point. As is the case for other kinds of ecological studies, the method of statistical analyses applied in many publications is not always appropriate. We note there are many statistical errors and misapplication of data analysis techniques. We reviewed 875 piosphere-related publications between 1915-2018 to find the common statistical methods and common statistical errors in the design of the study, data analyses, presentation of results, and interpretation of study findings. One-way ANOVA, multiple linear regression, Pearson correlation coefficient, permutational multivariate analysis of variance, canonical correspondence analysis, and mean were the most frequent statistical methods applied. Seventy-one common statistical errors in piosphere publications were found. The most common errors were not choosing the proper or appropriate statistical techniques, not checking the assumptions and diagnostics of statistical methods, partial and wrong interpretation of results, and not using informative figures and tables to help readers. Negligence to the proper application of statistics by researchers results in inaccurate interpretation and spurious conclusions. It is recommended researchers seek advice from statisticians at the early stages of research to save resources, time, and labor and to provide increased trust in recommendations and findings.