Chambers David W, Licari Frank W
Arthur A. Dugoni School of Dentistry, University of the Pacific, 2155 Webster Street, San Francisco, CA 94115, USA.
J Dent Educ. 2009 Mar;73(3):287-302.
Surveys are the most common form of data-based article published in the Journal of Dental Education. The apparent ease with which they are conducted and the seeming simplicity of reporting results mask significant issues in sample design and performing maximally useful analyses. Four concerns are discussed here. First, it is demonstrated that results are a function of who, when, and where responses are sampled, each source making independent contributions. Second, absolute sample size is shown to be the most significant factor affecting precision in surveys, and the numbers of schools, respondents, and other sources of variance can be chosen to minimize survey imprecision. Third, response rate typically has negligible effect on precision and an uncertain effect on accuracy (freedom from bias). A technique, sample saturation, is explained that can be used to protect, to some degree, surveys from the effects of bias. Finally, suggestions are offered for reporting survey results in a visually meaningful fashion, and an appeal is made that recommendations associated with surveys not be published unless they are grounded in both data and well-developed theory. This analysis references a previously published survey on competency-based dental education to illustrate methodological points in concrete terms.
调查是发表在《牙科教育杂志》上最常见的基于数据的文章形式。进行调查表面上的轻松以及报告结果看似的简单掩盖了样本设计和进行最有用分析方面的重大问题。这里讨论四个问题。首先，结果表明是由谁、何时以及在何处抽取回复样本决定的，每个来源都有独立的贡献。其次，绝对样本量被证明是影响调查精度的最重要因素，可以选择学校数量、受访者数量和其他方差来源，以尽量减少调查的不精确性。第三，回复率通常对精度影响可忽略不计，对准确性（无偏差）的影响不确定。解释了一种样本饱和技术，可用于在一定程度上保护调查免受偏差影响。最后，提供了以直观有意义的方式报告调查结果的建议，并呼吁除非基于数据和完善的理论，否则与调查相关的建议不应发表。本分析引用了之前发表的一项关于基于能力的牙科教育的调查，以具体说明方法要点。
