Preference Evaluation Research (PrefER) Group, Duke Clinical Research Institute, Duke University, Durham, NC, USA.
Preference Evaluation Research (PrefER) Group, Duke Clinical Research Institute, Duke University, Durham, NC, USA.
Value Health. 2019 Feb;22(2):157-160. doi: 10.1016/j.jval.2018.07.876. Epub 2018 Sep 27.
To develop a tool for testing internal validity of discrete choice experiment (DCE) data, deploy the program, and collect summary test results from a sample of active health researchers to demonstrate the practical utility of the tool in a wide range of health applications.
A previously developed Gauss program had been in use for testing internal validity. The program was translated to MATLAB and adapted, compiled, and deployed. Sixty-seven authors who had coauthored one or more published DCE studies between 2013 and 2016 were contacted by email; provided access to the tool, instructions, and an example data file; and invited to submit test summaries for tabulation.
Twenty-one researchers from 10 countries contributed test results from a total of 55 DCE data sets. Fifty-one studies included at least two out of a possible six tests. Attribute dominance was the most common test, and stability had the highest failure incidence. Only three summaries included a transitivity test, and no failures were detected.
It was possible to evaluate multiple internal validity checks for most data sets even when the experimental design did not explicitly include tests. Nevertheless, internal validity is rarely reported. Free availability of the tool for testing data quality could improve both reporting and more careful design of DCE studies to help validate and interpret stated preference data.
开发一种测试离散选择实验（DCE）数据内部有效性的工具，部署该程序，并从一组活跃的健康研究人员中收集汇总测试结果，以展示该工具在广泛的健康应用中的实际效用。
先前开发的 Gauss 程序已用于测试内部有效性。该程序已被翻译成 MATLAB 并进行了改编、编译和部署。2013 年至 2016 年间，曾合著过一篇或多篇已发表的 DCE 研究论文的 67 位作者通过电子邮件联系；提供了对工具、说明和示例数据文件的访问权限，并邀请他们提交测试摘要进行制表。
来自 10 个国家的 21 位研究人员从总共 55 个 DCE 数据集提交了测试结果。51 项研究至少包含六个测试中的两个。属性优势是最常见的测试，稳定性的失败发生率最高。只有三个摘要包括了一个传递性测试，没有发现失败。
即使实验设计没有明确包括测试，也可以对大多数数据集进行多项内部有效性检查。然而，内部有效性很少被报告。该工具可免费用于测试数据质量，可以提高报告的质量，并更仔细地设计 DCE 研究，以帮助验证和解释所提出的偏好数据。
