心理学可重复性危机两种根源的评估

发布时间：2018-05-29 18:59

本文选题：心理学可重复性危机 + 传统统计检验的局限　；参考：《心理与行为研究》2017年05期

【摘要】：心理学的可重复性危机有两大已知的根源:传统统计学中虚无假设显著性检验体系的局限,和心理学的学术传统中的弊端,本文以开放科学协作组2015年报告的数据为依据,试对这两个根源的影响作一粗略的估算。采用Goodman(1992)和Cumming(2008)提出的方法对传统统计体系所加诸于可重复性的限制加以分析后,估算的结果表明传统统计学体系的制约,虽然有举足轻重的影响,却远不能完全解释该报告中低至36%的可重复率,该报告所反映的状况,显然还另有重大的非统计学的根源。本文进一步用Ioannidis(2005)提出的模型对这类非统计学因素的影响加以分析。分析后得到的若干组人为偏差率和Ha真实概率的估算,表明在原来研究所获得的几乎清一色的阳性结果中,大约只有不到三分之一或更低的比例是真阳性,而且相当部分的阳性结果,可能由人为偏差所造成。这样的分析可比较具体地描述该类因素对当前可重复性危机的可能影响。
[Abstract]:There are two known causes for the repetitive crisis in psychology: the limitations of the test system of nothingness hypothesis in traditional statistics, and the drawbacks of the academic tradition of psychology. This paper is based on the data reported by the Open Scientific Cooperation Group in 2015. Try to make a rough estimate of the effects of these two sources. Using the methods proposed by Goodman (1992) and Cummingli (2008) to analyze the reproducible limitations imposed by traditional statistical systems, the estimated results show that the constraints of traditional statistical systems are important, although they have a significant impact on them. But far from fully explaining the report's low repetition rate of as low as 36 percent, the report clearly reflects another significant non-statistical source. In this paper, the influence of these non-statistical factors is analyzed by using the model proposed by Ioannidishe 2005. The estimates of several groups of artificial deviations and Ha's true probability obtained from the analysis indicate that, of the almost universal positive results obtained from the original study, only about 1/3 or less are true positive. And a considerable number of positive results, may be caused by human bias. Such an analysis can describe more specifically the possible impact of such factors on current repeatable crises.
【作者单位】：美国宾夕法尼亚印第安纳大学;
【分类号】：B841

【相似文献】