P-values: Interpretation and Methodology
P-values: Interpretation and Methodology
复制标题
P 值：解释和方法
DOI：
10.1080/00031305.1975.10479106
复制
发表时间：
1975
期刊：
影响因子：
--
通讯作者：
J. Pratt
中科院分区：
文献类型：
--
作者：
J. Gibbons;J. Pratt
执行任何假设检验的最常见的传统方法是选择用于拒绝的区域并形成拒绝规则，使得犯类型I错误的概率不超过被称为测试水平的某个预先选择的数字。然后，调查员报告观察结果在所选水平上是否“重要”。这一过程可能源于经典统计学中Neyman-Pearson理论的使用，在经典统计学中，测试的决策函数是这样确定的，即类型II错误的概率是受所选水平施加的条件的最小约束。这种测试构造方法避开了类型I和类型II错误的概率之间的相互关系的问题。然而，在许多情况下，重要程度的选择完全是任意的。特别是在非参数统计中，但在参数统计中，当零分布是离散的时，所选择的水平甚至可能无法达到。此外，在非参数统计中，通常没有足够的关于可选分布的信息，因此甚至可以一般地讨论类型II误差的可预测性。相反，决策函数是通过逻辑推理来选择的，或者根据研究假设，有时甚至通过数据来选择。假设检验的另一种方法目前正在获得广泛接受。这是一种做法，即报告在特定方向上观测显著的最小水平。这
The most common traditional method of carrying out any hypothesis test is to select a region for rejection and form a rejection rule such that the probability of committing a Type I error does not exceed some preselected number called the level of the test. Then the investigator reports whether or not the observations are" significant" at the chosen level. This procedure probably stems from the use of the Neyman-Pearson theory in classical statistics, where the decision function for the test is determined such that the probability of a Type II error is a minimum subject to the conditions imposed by the level selected. This method of test construction circumvents the problem of interrelationship between the probabilities of the Type I and Type II error. However, in many cases the choice of a significance level is completely arbitrary. In nonparametric statistics particularly, but also in parametric statistics when the null distribution is discrete, the chosen level may not even be attainable. Further, in nonparametric statistics, there is usually not sufficient information about alternative distributions so that the pr-obability of a Type II error can even be discussed in general. Rather, the decision function is selected by logical reasoning, or according to the research hypothesis, or sometimes even by the data. Another approach to hypothesis testing is currently attaining wide acceptance. This is the practice of reporting the smallest level at which the observations are significant in a particular direction. This