关键词: 心理学可重复性危机, 传统统计检验的局限, 心理学学术传统中的弊端, 备择假设真实概率, 人为偏差, 危机根源的评估

Abstract: The reproducibility crisis in psychology is known to have two roots, the root in the traditional statistical system of null hypothesis significance testing, and that in the academic tradition of psychology. This article was an attempt to crudely estimate the respective impacts of the two roots on the reproducibility crisis in psychology. The results reported by Open Science Collaboration (2015) were analyzed using the methods suggested by Goodman (1992) and by Cumming (2008) to roughly estimate the limiting influence on reproducibility imposed by the traditional system of statistics. The estimated limiting influence, although quite notable, appears to be far short of being able to account for the reproducibility rate as low as 36% indicated by the report, suggesting that factors other than the traditional system of statistics have played a tremendous role in the crisis. The model proposed by Ioannidis (2005) was adopted to analyze the possible impacts of factors other than the traditional system of statistics, and possible ranges of the joint impact of bias and the probability of true alternative hypotheses were extrapolated。 The analysis led to estimates indicating that, of all original positive results, only no more than one third, and probably even less, was true positive, and a considerable portion of these positive results was caused by bias. These results may help explicate how these factors are likely to contribute to the current crisis.

Key words: the reproducibility crisis in psychology, limitations of traditional system of statistical testing, flaws in the academic tradition of psychology, the probability of true alternative hypotheses, bias, evaluation on the roots of the reproducibility crisis