Statistical Significance,<i>p</i>-Values, and the Reporting of Uncertainty

The use of statistical significance and p-values has become a matter substantial controversy in various fields using methods. This gone as far some journals banning the indicators for significance, or even any reports p-values, and, one case, mention confidence intervals. I discuss three issues that have led to these often-heated debates. First, argue many cases, do not answer questions primary interest. Such typically involve making (recommendations on) decisions under uncertainty. In point estimates measures uncertainty form intervals better, Bayesian intervals, are often more informative summary statistics. fact, presence absence is essentially irrelevant, including them discussion may confuse at hand. Second, there also cases where testing null hypotheses natural goal reasonable appropriate conclude general counterproductive. Third, overemphasis empirical work on abuse p-hacking publication bias. pre-analysis plans replication studies, combination with lowering emphasis help address problems.