中文摘要：

人们熟知的零假设显著性检验,受到一次次质疑与辩护,地位并未动摇,报告检验结果仍然是统计分析的习惯做法。不过,其局限性促使研究者探寻更多的统计方法如区间估计、效应量分析、检验力分析等。本文先介绍假设检验与置信区间的关系;然后讨论检验力与两类错误率和效应量的关系;最后在理顺上述统计方法的基础上,提供一个可操作的统计分析流程。

英文摘要：

Although the null hypothesis significance testing（NHST） has long been argued about,it is still frequently adopted in psychological researches.Reporting the test results is still the usual practice.The limitations of NHST recognized through the arguments,however,urged the researchers to explore and apply other statistical approaches such as interval estimation,effect size analysis,power analysis. We summarized the relations between NHST and confidence interval,before dealing with the relation of test power,with two types of error rate and the effect size.Under certain research conditions,test power is related to effect size,sample size and significant level（α）.Power is higher when the effect size is larger,when the sample size is larger and when a is higher. We provided an operational statistical analysis procedure for the research involving NHST.The first step was a prior power analysis before sampling through which we could calculate the necessary sample size according to the acceptable a and power as well as the smallest effect size that we expected to detect.In the second step,we reported the result of NHST and confidence interval after the sampling.The third step was to analyze the effect size and decide whether or not we needed a post-hoc power analysis based on the result of NHST.It would be better to report the confidence interval of the effect size.When the result of NHST was significant,we could make a research conclusion according to the magnitude of the effect size.The effect of the independent variable on the outcome was insignificant when the effect size was small even if the result of NHST was significant.When the result of NHST was not significant,there were two kinds of situations： （1） if the effect size was small,accepted the null hypothesis and made a conclusion that the independent variable did not affect the outcome. （2）if the effect size was medium or large,we needed a post-hoc power analysis.That is,we calculated the we actual power of the research：（2.1） we accepted