Planned Missing Research Designs: Power and Validity of Planned Missing Data Designs in Longitudinal Research

计划缺失研究设计：纵向研究中计划缺失数据设计的功效和有效性

• 批准号: 1053160
• 负责人: Wei Wu
• 金额: $ 40万
• 依托单位: University of Kansas Center for Research Inc
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-06-15 至 2016-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1053160&HistoricalAwards=false
• 关键词: Planned; Missing; Research; Designs; Power

This project will develop and evaluate a research methodology known as a "planned missing research design." This statistical technique enables scientists who conduct longitudinal studies to obtain higher quality results (greater statistical certainty and/or broader ability to generalize findings) using the same amount of research resources. This technique has been used with great effectiveness in certain specific contexts in the past. The proposed project will provide guidelines and procedures that will enable researchers to use the technique in all areas of developmental science in which a study evaluates how participants change over time.Planned missing data designs have tremendous potential to streamline how we conduct basic research in the social, behavioral, and educational sciences. Strong statistical theory supports the idea that carefully planned patterns of intentionally missing information will not bias or weaken the accuracy of research conclusions. When properly constructed a planned missing design can save 33% or more of the costs associated with data collection. They can also reduce the fatigue and demand on participants which can increase the overall quality of the data that is collected from each participant. In fact, planned missing research designs have been recommended for decades as efficient ways to collect expensive data (e.g., one-on-one assessments) and time-intensive data (e.g., large questionnaire protocols and repeated measures). Planned missing data designs can be used to manage cost, improve data quality, reduce fatigue of participants and test-retest effect, as well as increase statistical power to detect effects of interest. Statisticians are certain of these benefits; researchers and stakeholders, however, remain skeptical. Outside the ideal conditions that statisticians espouse for such designs, are they practical for the conditions faced in real-world research? Statistical proofs are just a starting point for answering important questions about the practical nature of these designs. This research will provide compelling guidance to researchers for using planned missing research designs in social, behavioral, and educational research.To validate their utility for practical researchers, planned missing designs will be thoroughly studied under varying conditions of real-world data. The goal of this sponsored research is to examine precisely under what conditions such designs can be useful and the conditions in which caution would be warranted. Previous studies on planned missing data designs have all been small scale simulation studies that demonstrate proof of concept but do not provide practical guidance to researchers. This funded project will conduct a series of well-conceived large-scale Monte Carlo simulation studies to explore the utility of planned missing data designs under a wide variety of conditions including extreme data conditions and complex longitudinal designs. This project will vary the critical features of planned missing designs that will inform researchers on when, how, and why to use planned missing elements in their research designs. The results of this project fully detail the boundaries of when this design is useful and when it should be avoided or modified. By so doing, the results of this work will be paradigm shifting for research across the social, behavioral, and educational sciences because real-world researchers and stakeholders will now understand their utility and applicability. As a result, this funded project will impact and change the future of best-practice methodology in all these research areas.

这个项目将开发和评估一种被称为“计划缺失研究设计”的研究方法。“这种统计技术使进行纵向研究的科学家能够使用相同数量的研究资源获得更高质量的结果（更大的统计确定性和/或更广泛的概括发现的能力）。过去在某些特定的情况下，这种技术已经被非常有效地使用。拟议的项目将提供指导方针和程序，使研究人员能够在发展科学的所有领域使用该技术，其中研究评估参与者如何随着时间的推移而变化。计划缺失数据设计具有巨大的潜力，可以简化我们如何进行社会，行为和教育科学的基础研究。强有力的统计理论支持这样一种观点，即精心策划的故意缺失信息的模式不会使研究结论产生偏见或削弱其准确性。如果构造得当，计划缺失设计可以节省33%或更多的数据收集相关成本。它们还可以减少参与者的疲劳和需求，从而提高从每个参与者收集的数据的整体质量。事实上，几十年来，计划缺失的研究设计一直被推荐为收集昂贵数据的有效方法（例如，一对一评估）和时间密集型数据（例如，大型问卷调查协议和重复测量）。计划缺失数据设计可用于管理成本，提高数据质量，减少参与者的疲劳和重测效应，以及增加检测感兴趣效应的统计功效。统计学家确信这些好处;然而，研究人员和利益相关者仍然持怀疑态度。在统计学家支持的理想条件之外，这些设计对于现实世界研究中面临的条件是否实用？统计证明只是回答关于这些设计的实际性质的重要问题的起点。这项研究将为研究人员在社会、行为和教育研究中使用计划缺失研究设计提供令人信服的指导。为了验证它们对实际研究人员的效用，将在不同的现实数据条件下彻底研究计划缺失设计。这项赞助研究的目标是精确地研究在什么条件下这种设计是有用的，以及在什么条件下需要谨慎。以前的研究计划缺失数据设计都是小规模的模拟研究，证明了概念，但不提供实际指导的研究人员。该资助项目将进行一系列精心设计的大规模蒙特卡罗模拟研究，以探索在各种条件下（包括极端数据条件和复杂纵向设计）计划缺失数据设计的实用性。该项目将改变计划缺失设计的关键特征，这将告知研究人员何时，如何以及为什么在他们的研究设计中使用计划缺失元素。该项目的结果充分详细说明了这种设计何时有用以及何时应该避免或修改的界限。通过这样做，这项工作的结果将为社会，行为和教育科学的研究带来范式转变，因为现实世界的研究人员和利益相关者现在将了解它们的效用和适用性。因此，该资助项目将影响和改变所有这些研究领域的最佳实践方法的未来。