从不显著结果中提取信息的方法：原理及其实现

In empirical studies, researchers need to prove that the null hypothesis is true. Unfortunately, non-significant results, when p > .05, cannot distinguish the absence of evidence from the evidence of absence. Even though, p > .05 was always mistakenly interpreted as the evidence of absence. To address this issue, researchers have begun employing equivalence test, Bayesian estimation, and Bayes factor to draw more informative conclusions from non-significant result. Specifically, equivalence test enables us to evaluate the null effect by testing whether the detected effect match the smallest effect size of interest (SESOI). Likewise, the Bayes estimation help us to accept or refuse null effect by estimating the overlap between highest density intervals (HDI) of posterior distribution and region of practical equivalence (ROPE). Also, Bayes factor enable us to assess the relative strength of evidence for null hypothesis and alternative hypothesis. Using equivalence test or Bayesian estimate to make inference requires researcher to define a SESOI or ROPE beforehand, and the SESOI or ROPE are usually based on theories or data from previous studies. In this Chinese tutorial, we explained how equivalence test and Bayesian estimate work and illustrated how to apply equivalence test, Bayesian estimate and Bayes factor through a worked example. By doing so, we hope that Chinese researchers can apply those statistical techniques to interpret non-significant results in a better way, and therefore promote the accumulation of empirical knowledge."