Statistics Reform: Practitioner’s Perspective

It is generally accepted that one of the main causes of the replication crisis in scientific research is some of the most commonly used statistical methods, such as null hypothesis significance testing (NHST). This has prompted many scientists and statisticians to call for statistics reform. As a practitioner in the fields of hydraulics and measurement science, the author used statistical methods extensively in many environmental engineering and hydrological survey projects. The author strongly concurs in the need for statistics reform. This paper offers a practitioner’s perspective on statistics reform. In the author’s view, some statistical methods are good and should withstand statistics reform; some are bad and should be abandoned and removed from statistics textbooks and computer software packages. This paper focuses on two statistical methods derived from the _t_-distribution: the two-sample _t_-test and the _t_-interval method for measurement uncertainty calculation. We show why these two methods should be abandoned. We recommend using descriptive statistic analysis as an alternative to the two-sample _t_-test and an unbiased estimation method as an alternative to the _t_-interval method for measurement uncertainty calculation. Two examples are provided to demonstrate the recommended alternatives.