Ottenbacher K J
School of Health Related Professions, State University of New York at Buffalo 14214, USA.
J Dev Behav Pediatr. 1995 Jun;16(3):177-82.
The purpose of this investigation was to examine quantitative methods used to determine reliability in developmental research. Procedures used to compute reliability estimates in 30 studies published in three developmental journals were examined. Four types of reliability studies were identified and analyzed. These included interrater reliability, stability (test-retest and intrarater reliability), equivalence reliability, and internal consistency. Interrater reliability investigations were the most frequently reported in the developmental literature reviewed (45%). The Pearson product moment correlation (r) was the most commonly reported reliability statistic. The findings reveal that researchers in developmental pediatrics frequently analyze reliability data using the Pearson product moment correlation and interpret the results as indicating consensus (agreement) among raters or across instruments. The Pearson product moment correlation (r) provides information on covariation among variables but does not indicate agreement. Thus, the findings suggest that developmental researchers may be misinterpreting the statistical results of reliability investigations. The argument is made that the intraclass correlation coefficient (ICC) is a more appropriate method of analysis when the purpose of the research is to examine consensus.
本调查的目的是检验用于确定发展性研究中可靠性的定量方法。对发表在三种发展性期刊上的30项研究中用于计算可靠性估计值的程序进行了检查。确定并分析了四种类型的可靠性研究。这些包括评分者间信度、稳定性（重测信度和评分者内信度）、等效性信度和内部一致性。在回顾的发展性文献中，评分者间信度调查的报告最为频繁（45%）。皮尔逊积差相关系数（r）是最常报告的可靠性统计量。研究结果表明，发育儿科学领域的研究人员经常使用皮尔逊积差相关系数来分析可靠性数据，并将结果解释为表明评分者之间或不同工具之间的一致性（一致性）。皮尔逊积差相关系数（r）提供了变量间协变的信息，但并不表明一致性。因此，研究结果表明，发展性研究人员可能误解了可靠性调查的统计结果。有人认为，当研究目的是检验一致性时，组内相关系数（ICC）是一种更合适的分析方法。
