Thou Shalt Not Reject the $\textit{P}$-value.

Since its debut in the 18th century, the $\textit{P}$-value has been an integral part of hypothesis testing based scientific discoveries. As the statistical engine ages, questions are beginning to be raised, asking to what extent scientific discoveries based on a $\textit{P}$-value ($\textit{e.g.}$, the practice of drawing scientific conclusions relying on the fact that the $\textit{P}$-value is smaller than an artificially determined threshold, for example, that of 0.05) are reliable and reproducible, and the voice calling for adjusting the significance level of 0.05 or banning the $\textit{P}$-value has been increasingly heard. Inspired by these questions, we inquire into the useful roles and misuses of the $\textit{P}$-value in scientific studies. We attempt to unravel the associations between the $\textit{P}$-value, sample size, significance level, and statistical power. For common misuses and misinterpretations of the $\textit{P}$-value, we provide modest recommendations for practitioners. Additionally, we review, with a comparison, Bayesian alternatives to the $\textit{P}$-value, and discuss the advantages of meta-analysis in combining information, reducing bias, and delivering reproducible evidence. Taken together, we argue that the $\textit{P}$-value underpins a useful probabilistic decision-making system, provides evidence at a continuous scale, and allows for integrating results from multiple studies and data sets. But the interpretation must be contextual, taking into account the scientific question, experimental design (including the sample size and significance level), statistical power, effect size, and reproducibility.