GB/T 4882-2001 in English

本标准在假定观测值相互独立时，对决定分布是否为正态的假设应否被拒绝的方法和检验，给出了一个导引。
当对观测值是否服从正态分布存在疑问时，使用偏离正态分布的检验是有用的，甚至是必须的。利用t检验检查一个随机观测样本的均值是否偏离给定的理论值，就是这种情况的一个例子。然而，在稳健方法(即观测值的真实的概率分布不是正态时。结论仅有轻微的变化)的情况下，偏离正态分布的检验并不是非常必要的。
涉及基于正态性假设的统计方法时，也并非严格地必须使用这样一个检验。观测值的正态分布可能是完全没有疑问的。可以是理论的(如物理的)原因构成了这个假设，也可以是根据先验信息接受了这个假设。
本标准中偏离正态分布的检验是针对非分组的原始数据，而不是分组数据。检验也不适用于截尾数据。
本标准中偏离正态分布的检验可以应用于观测值，也可以应用于它们的函数，如取对数、平方根等。当样本容量小于8时，偏离正态分布的检验效果是非常差的。因此，本标准限制样本量至少为8。

Scope

1.1 This standard gives guidance on methods and tests for determining whether the assumption of a normal distribution should be rejected when the observed values are assumed to be independent of each other. 1.2 When there is doubt as to whether observations follow a normal distribution, it is useful, even necessary, to use a test for deviations from the normal distribution. An example of this is the use of the t-test to check whether the mean of a sample of random observations deviates from a given theoretical value. However, in the case of robust methods (i.e. the conclusions change only slightly when the true probability distribution of the observations is not normal), testing for deviations from the normal distribution is not very necessary. 1.3 Nor is it strictly necessary to use such a test when it comes to statistical methods based on assumptions of normality. The normal distribution of the observed values may be completely unquestionable, there may be theoretical (eg physical) reasons for this assumption, or the assumption may be accepted based on prior information. 1.4 The test of deviation from normal distribution in this standard is for non-grouped raw data, not grouped data. The test also does not apply to censored data. 1.5 The test of deviation from normal distribution in this standard can be applied to observed values, and can also be applied to their functions, such as taking logarithm, square root, etc. 1.6 When the sample size is less than 8, the test effect of deviation from the normal distribution is very poor. Therefore, this standard limits the sample size to at least 8.