| 研究结果三大评价指标的困惑与思考 |
| |
| 引用本文： | 张力为.研究结果三大评价指标的困惑与思考[J].北京体育大学学报,2002,25(2):211-217. |
| |
| 作者姓名： | 张力为 |
| |
| 作者单位： | 北京体育大学,北京,100084 |
| |
| 摘 要： | 讨论了显著性、效果量和原因量在研究设计和统计分析中的意义与作用.指出,不应将显著性检验中规定的0.05水平绝对化,对刚刚高于或刚刚低于此水平的研究结果,应持谨慎或宽容的态度.同时,根据Pollard(1993)的分析,显著性水平提供的是I型错误的条件先验概率和I型错误的整体先验概率的上限,而不是I型错误的条件后验概率.因此,应注意先验概率和后验概率的区分.效果量与显著性有本质区别,研究者不仅应关注和报告显著性,也应关注和报告效果量.多大的效果量才有实际意义,取决于研究者对特定研究领域特定研究问题的专业判断.研究者进行研究设计时,还必须考虑效果量和原因量的关系问题,即因果效能问题,设法用最小的代价产生最大的效益.
|
| 关 键 词： | 统计学 显著性 效果量 原因量 |
| 文章编号： | 1007-3612(2002)02-0211-07 |
| 修稿时间： | 2001年12月4日 |
Dilemma and Consideration of Three Indexes of Research Result Evaluation |
| |
| ZHANG Li,wei.Dilemma and Consideration of Three Indexes of Research Result Evaluation[J].Journal of Beijing Sport University,2002,25(2):211-217. |
| |
| Authors: | ZHANG Li wei |
| |
| Abstract: | The meaning and function of significance, effect size and cause size in research design and statistical analysis were discussed. It was pointed out that the level of .05 in the significant test should not be considered as an absolute criterion. The result of just above or below the level of .05 should be considered with caution or tolerance. Meanwhile, according to Pollard (1993), what significant level provides is the conditional prior probability and the upper bound of the whole prior probability, rather than the conditional posterior probability of a Type I error. Therefore, the prior probability and posterior probability should be clearly distinguished in a significant test. It was also pointed out that the effect size and the significance are totally different concepts. Researchers should pay close attention to and report not only the significance but also the effect size in their research reports. How large effect size could be taken as having practical meaningfulness depends on the researchers' professional judgment on specific problems in their fields. The relationship between effect size and cause size, i.e. causal efficacy, should also be considered. Maximum benefit should be tried to gain at minimum cost. |
| |
| Keywords: | statistics significance effect size cause size |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
