多维空间全面分析测量可靠性及其误差成分——多因素方差分析在估价可靠性中的应用

Reliability of Overall Evaluation and Its Variance-Components in Multi-dimensional Analysis -Application of Multi-Factor ANOVA in Reliability Evaluation

分析比较认为,多因素方差法可从多维空间来分析具有多层面、多种误差的测验观测数,可鉴别出受试者之间(S)、测验之间(D)和测试者之间(T)的误差,以及受试者与各天(次)测验(S·D)之间,受试者与测试者之间(S·T),测试者与各天(次)测验之间(D·T)及受试者、测试者与各天(次)之间(S·D·T)的交互影响(差异);通过误差成分分析,不仅可以求出各误差成分对可靠性影响的程度,而且还可以求出可靠性系数。

The purpose of this study is to utilize multi-factor ANOVA to analyze the reasons of variance, variance-components and their influence on reliability and to discuss the calculation and changes of reliability coefficient under the situation of various variance-components coexistence. It was concluded that multi-factor ANOVA could be used to analyze data with multi-facets and multi-variances, to distinguish the variances amoug subjects, days, and conductors, as well as the mutual influence between subjects and days, subjects and conductors, days and conductors, and that between all subjects, days and conductors. With analysis of variance components, not only the influence level of each component to reliability could be evaluated, but also the reliability coefficient could be calculated. It was also found that multi-factor ANOVA was more ail-roundly and accurately in reliability of evaluation than bi-serial correlation, one-way and two-way ANOVA.