Science, pseudoscience and the need for practical knowledge.
Science, pseudoscience and the need for practical knowledge.
复制标题
科学、伪科学和实践知识的需要。
DOI：
10.1111/j.1360-0443.2008.02290.x
复制
发表时间：
2008
期刊：
影响因子：
--
通讯作者：
Gorman,DennisM
中科院分区：
文献类型：
--
作者：
Gorman,DennisM
在她最近对图米等人的评论中，格雷厄姆[2]在对酒精风险管理计划的评估中认为，改变研究人员设计社区预防计划评估的方式，会更好地满足社区试图处理与在有执照的场所饮酒有关的问题的实际需要。具体来说，她认为，如果研究人员在很长一段时间内测量大量的结果变量，并使用0.10和0.20的P值评估干预计划对这些多个结果的影响，然后可以确定“可能有效”或“有希望”的干预措施。大部分药物和酒精预防领域已经朝着减少科学研究的方向发展。格雷厄姆推荐的严谨以学校为基础的项目的评估，例如，使用单尾检验和P值大于0.05的常规显著性检验，沿着的做法，如多个亚组分析，改变研究报告的结果变量，和事后样本细化[3-5]。使用这种数据分析实践的评估几乎总是产生一些积极的计划效果，这些孤立的统计学显著结果足以满足美国药物滥用和精神卫生服务管理局和美国教育部等机构的证明或循证标准[3，5]。然而，使用这种做法的问题是，它们创造了一种伪科学的研究文化，这种文化是非批判性的，寻求验证（而不是伪造）其假设，很少拒绝任何与计划有效性有关的质疑，忽视和/或解释不利的数据[6，7]。进行的每一次评估都只是为了确认干预方案是有效的。
In her recent commentary on Toomey et al.’s [1] evaluation of the alcohol risk management program, Graham [2] contends that the practical needs of communities trying to deal with problems related to drinking in licensed premises would be better served by a change in the way in which researchers design evaluations of community prevention programs. Specifically, she argues that if researchers were to measure a large number of outcome variables over a long period of time and assess the effects of intervention programs on these multiple outcomes using P values of 0.10 and 0.20, then interventions that are of ‘probable effectiveness’ or ‘promising’could be identified.Much of the drug and alcohol prevention field has already moved in the direction of reduced scientific rigor that Graham recommends. Evaluations of school-based programs, for example, use one-tailed tests and P values greater than 0.05 routinely in significance testing, along with practices such as multiple subgroup analysis, altering outcome variables from research report to research report, and post hoc sample refinement [3–5]. Evaluations that use such data analysis practices nearly always produce a few positive program effects, and these isolated statistically significant results are sufficient to meet the proven or evidence-based criteria of agencies such as the US Substance Abuse and Mental Health Services Administration and US Department of Education [3, 5]. However, the problem with the use of such practices is that they create a pseudoscientific research culture that is non-critical, seeks verifications (rather than falsification) of its hypotheses, seldom rejects any conjectures pertaining to program effectiveness and ignores and/or explains away unfavorable data [6, 7]. Every evaluation conducted simply serves to confirm that the intervention program is effective.