PsychStatsWiki

本页面只读。您可以查看源文件，但不能更改它。如果您觉得这是系统错误，请联系管理员。
==== 效应大小的用途 ====
=== Uses of effect sizes ===
  * 处理效应越明显，越容易被侦察到
  * The more pronounced the treatment effect, the more likely it is to be detected
  * 两总体差异越大，即处理效应越大，处理后的总体中位于临界内的部分较多，阴影区域较大，从而效力较高；右边的图中两总体差异较小，处理后总体位于临界区域的面积，即阴影区域较小，效力较低
  * The larger the difference between the two aggregates, i.e., the larger the treatment effect, the more of the treated aggregate lies within the critical area, the larger the shaded area, and thus the higher the validity; in the figure on the right the difference between the two aggregates is smaller, and the treated aggregate lies in the critical area, i.e., the shaded area is smaller, and the validity is lower
  * 一个统计检验可能显著但是效应太小以至于没有实际意义
  * a statistical test may be significant but the effect is so small as to be meaningless
  * 当不显著时，效应较大能说明一定的问题
  * When it is not significant, a large effect can mean something

==== 效应大小的评价 ====
=== Evaluation of effect size ===
  * d=(M-μ)/σ

^  效应大小 (effect size)  ^  评价 （estimate）  ^
|  0<d<0.2   |  小的效应(mean difference<.2 SD)  | 
|  0.2<d<0.8  |  中等效应(mean diffenrence around.5 SD)  | 
|  d>0.8   |  大的效应(mean difference >.8 SD)   |

  * 拒绝H0时，大的效应显示这是一个重要的效应，小的效应提示研究者应该另辟蹊径
  * When H0 is rejected, a large effect shows that it is a significant effect, and a small effect suggests that the researcher should find another way to do it
  * 接受H0时，差异不显著是因为效力低，效应大说明应该在大样本中重复研究
  * When H0 is accepted, the difference is not significant because of low validity and the large effect suggests that the study should be repeated in a large sample

==== 效力和效应大小的区别 ====
=== Difference between power and effect size ===
  * 效力是侦察效应的能力，是统计检验的特性
  * Power is the ability to detect effects and is a characteristic of statistical tests
  * 效应大小是两个分布之间重叠程度的大小，是分布的特性
  * Effect size is the magnitude of overlap between two distributions and is a characteristic of the distribution