A psychometric toolbox for testing validity and reliability
A psychometric toolbox for testing validity and reliability
复制标题
DOI：
10.1111/j.1547-5069.2007.00161.x
复制
发表时间：
2007-01-01
影响因子：
3.4
通讯作者：
Kostas-Polston, Elizabeth
中科院分区：
文献类型：
--
作者：
DeVon, Holli A.;Block, Michelle E.;Kostas-Polston, Elizabeth
目的：回顾信度和效度的概念，举例说明这些概念在护理研究中的应用，为提高心理测量工具的可靠性提供指导，并报告护理期刊编辑对将心理测量数据纳入论文的建议。方法：以效度、信度和心理测量学为关键词对CINAHL、MEDLINE和PsycINFO数据库进行检索。如果护理研究论文发表于最近5年内，采用定量方法，并报告了心理测量特性的统计证据，则符合纳入条件。报告强烈的心理测量的性质的工具以及那些支持证据很少的心理测量的健全。研究结果：报告经常指出内容效度，但有时研究的评审专家少于5人。标准的效度很少被报道，并且在标准的测量中发现了错误。结构效度仍然未被充分报道。大多数报告表明内部一致性信度（alpha），但很少有报告包括稳定性的信度测试。当重测信度被断言时，时间间隔和相关性通常不包括在内。结论：通过设计和减少测量中的非随机误差来规划心理测量将增加工具的信度和效度，并增加研究结果的强度。由于样本量小、设计差或缺乏资源，可能会出现效度低报。在文献中，缺乏关于心理测量特性的信息和误用心理测量测试是很常见的。
Purpose: To review the concepts of reliability and validity, provide examples of how the concepts have been used in nursing research, provide guidance for improving the psychometric soundness of instruments, and report suggestions from editors of nursing journals for incorporating psychometric data into manuscripts.Methods: CINAHL, MEDLINE, and PsycINFO databases were searched using key words: validity, reliability, and psychometrics. Nursing research articles were eligible for inclusion if they were published in the last 5 years, quantitative methods were used, and statistical evidence of psychometric properties were reported. Reports of strong psychometric properties of instruments were identified as well as those with little supporting evidence of psychometric soundness.Findings: Reports frequently indicated content validity but sometimes the studies had fewer than five experts for review. Criterion validity was rarely reported and errors in the measurement of the criterion were identified. Construct validity remains underreported. Most reports indicated internal consistency reliability (alpha) but few reports included reliability testing for stability. When retest reliability was asserted, time intervals and correlations were frequently not included.Conclusions: Planning for psychometric testing through design and reducing nonrandom error in measurement will add to the reliability and validity of instruments and increase the strength of study findings. Underreporting of validity might occur because of small sample size, poor design, or lack of resources. Lack of information on psychometric properties and misapplication of psychometric testing is common in the literature.