==== 事后检验（Post-hoc test） ====
  - ANOVA无法说明哪个备择假设得到支持；
     * The ANOVA cannot indicate which alternative hypothesis is supported.
  - 要进行事后检验以说明各组差异在什么地方；
     * Post-hoc tests are needed to show where the groups differ.
  - 事后检验是比较每一个组和另一个处理组，一次比较两个，这称为成对比较。
     * Post-hoc test compares each group to another treatment group, two at a time; this is called a pairwise comparison.

----
=== 族系（familywise）误差 ===
  - 每一个比较都是一个单独的假设检验，每一个都有犯I类错误的风险。所以，比较对数越多，作结论的风险越大。即容易发现实际不存在的差异。这称为实验导致的（experimentwise）alpha水平或族系（familywise）误差；
    *  Each comparison is a separate hypothesis test, each with the risk of making a Type I error. So the more pairs of comparisons there are, the greater the risk of making a conclusion. That is, it is easy to find differences that do not actually exist. This is called the **experimentwise** alpha level or **familywise** error.
  - 多次比较后犯I类错误的概率(Probability of making a Type I error after multiple comparisons)为**α<sub>EW</sub>=1-(1-α）<sup>c</sup>**，其中c为比较的次数(c=Number of comparisons)。

----
=== Tukey's HSD检验 (Tukey's HSD test) ===
  - 此检验各组要有相同的样本容量;
    * This test requires the same sample size for each group.
  - 可以计算出单一的值确定处理均值间的最小差异，考查此差异在统计上是否显著;
    * A single value can be calculated to determine the smallest difference between treatment means, to see if the difference is statistically significant.
  - {{::hsd公式.png?200|}}
  - q可以查表得到。
    * q score can be obtained by looking up the table.

----
=== Scheffe检验 (Scheffe test) ===
  - 适用于n不相等的情况;
    * Applies when n is not equal.
  - 降低I类错误的风险，提高II类错误的风险。
    * Reduces the risk of Type I errors and increases the risk of Type II errors.