领导干部结构化面试信度的多元概括化理论分析

Reliability Analysis of Structured Interview: A Multivariate Generalizability Theory Approach

本研究尝试运用多元概括化理论对北京市某区副处级干部准入资格结构化面试测评数据进行测量信度分析,为提高领导干部考试与测评工作科学化水平提供了有益的实证依据。主要结论有:(1)本次结构化面试难度适中,区分度较高;(2)各测评要素及合成分数的类信度系数均较高,合成分数的测量信度高于单个测评要素的测量信度;(3)各测评要素及合成分数的类信度系数随着考官数量的增加而增加,且从确保信度和降低成本考虑,考官数量以5-9位为宜;(4)在这次面试测评中,各项测评要素间的相关系数较高,这为目前在选拔面试中将各项测评要素得分进行合成提供了依据,说明用合成分数计算总分具有一定的合理性。

This paper attempted to apply the MGT in analyzing the reliability of structured interview. The results showed that, （1） the structured interview is moderately difficult and highly discriminated, （2） the reliabilities of both single and composite measurement aspects arc high, （3） the reliability of structured interview increases with the number of assessors （five to nine assessors are appropriate when balancing reliability and cost）, and （4） measurement aspects in structured interview are highly interrelated, which makes composite score reasonable. The implications for improving the present practice of selection interview are discussed.