====== 假设检验中的两类错误======
==== Two type of errors in hypothesis testing ====


^                       ^   没有效应，H<sub>0</sub>正确     ^    存在效应，H<sub>0</sub>错误      ^
^  拒绝H<sub>0</sub>     |  I类错误               |     正确                  |
^  接受H<sub>0</sub>     |  正确                  |      II类错误             |

^                       ^   false effect, H<sub>0</sub>is correct     ^    real effect, H<sub>0</sub>is wrong      ^
^  Reject H<sub>0</sub>     |  Type I error               |     Correct                  |
^  Accept H<sub>0</sub>     |  Correct                  |      Type II error             |

两种错误，即：

The two types of errors are:

  * α 错误，第一类错误（type I error ） ： 拒绝H<sub>0</sub>时所犯的错误，它原本成立，但被拒绝。即侦察到本不存在的差异。\\

  * α error(type I error): Rejecting the originally correct H<sub>0</sub>. That is to say, discovering differences that do not originally exists.\\

  * β 错误，第二类错误（type II error） ： 接受H<sub>0</sub>时所犯的错误，它原本不成立，但被接受。即未能侦察到存在的差异。\\

  * β error(type II error): Accepting the originally wrong H<sub>0</sub>. That is to say, not discovering differences that originally exists.\\

====== 显著性水平 ======
====level of significance====

在科学研究中，我们常常采取保守的策略，设定一个可接受的α水平来尽量减少I类错误。该α水平称为显著性水平。它规定了当虚无假设正确时，样本结果非常不可能出现的概率值。心理学研究中，α通常为0.05。
((《心理学名词》：显著性水平是指在统计推断中，观察值或假设检验统计量落入估计区间或拒绝域的概率。是统计推断错误的概率指标。))
((APA Dictionary:  In SIGNIFICANCE TESTING, a fixed probability of rejecting the NULL HYPOTHESIS of no effect when it is in fact true. It is set at some value, usually .001, .01, or .05, depending on the consequences associated with making a TYPE I ERROR. When a particular effect is obtained experimentally, the PROBABILITY LEVEL (p) associated with this effect is compared to the significance level. If the p value is less than the a level, the null hypothesis is rejected. Small p values suggest that obtaining a statistic as extreme as the one obtained is rare and thus the null hypothesis is unlikely to be trae. The smaller the a level, the more convincing is the rejection of the null hypothesis.))

  * In scientific research, we often adopt a conservative strategy to set an acceptable α level to minimize class I errors. This α level is called significance level. It specifies the probability that the sample result is very unlikely when the nothingness hypothesis is correct.In psychological research, it's usually 0.05.