【作者】 杨志明； 陈海平；

【Author】 Yang Zhiming~1 Chen Haipin~2 1 Educational Testing Service,Princeton,NJ,USA, 2 School of Psychology,Beijing Normal University,Beijing,100875

【机构】 Educational Testing Service； 北京师范大学心理学院；

【摘要】 常模(Norm)是将测验的原始分数(raw score)转化为有意义的量表分数(scale score)的工具或桥梁。常模质量的高低决定着人事选拔工作的成败。但要研发出高质量的常模,需要克服许多困难,接受许多挑战。这些挑战包括代表性不足、实证性不强、稳定性较差、平滑性不好和可比性欠缺等问题。其中,代表性不足指的是被试样本对目标总体的代表性不足、以及题目样本对测量内容全域或结构维度的代表性不足的问题。实证性不强指的是很多人事测评量表没有收集到足够的、独立的和实证的效度证据问题,也包括没有有效利用典型行为样本来标定常模的单位和界定分数的范围等问题。稳定性较差是指因题目记分设计不当或问卷过短等问题而导致的信度太低、误差太大的问题。平滑性不好指的是常模出现了天花板效应、地板效应、分数突变、以及分数逆转等问题。可比性欠缺指的是分数之间没能实现等值或垂直链接的问题。针对这些问题,安格夫(Angoff,1971,1984)曾从经典测验理论(C1assical Testing Theory,CTT)的角度作过系统论述。后来,葛萨齐(Gorsuch,1983,1985)从统计学的角度提出了"连续型常模"(Continuous Norming)的常模研发思路,该思路使得当代心理测量量表的研发水平上升到了一个崭新的高度。近年来,随着项目反应理论(Item Response Theory,IRT)的发展,居于IRT真分数等值方法的常模研发技术又有了新的进展。本文利用这些最新成果,探讨了应对常模研发中这些挑战的若干对策,为研发高质量的测验常模、提高人们鉴别量表质量的水平提供了一些建议。更多还原

【Abstract】 Norm is a conversion table that matches a raw score to a meaningful scaled score.It is essential for any assessment scale to have a high quality norm.A poor norm leads to an invalid assessment and thus fails to recruit the most appropriate candidates for prospective employers.To develop a high quality norm for an assessment scale,several challenges must be faced and solutions should be explored.Typically,if the sample of test takers inadequately represents the target population or if the sample of items fails to represent the desired contents or domains,the resulting norms will be inaccurate.Lack of independent validity evidences,low reliability, un-smoothness of the raw score to scale score conversion lines,and shortage of test equating or linking are other common challenges in norm development To explore solutions to fix the problems in norm development,Angoff(1971,1984)analyzed several norm development methods based on Classical Testing Theory((JIT).Gorsuch(1983,1985)proposed a continuous norming method for norm development.The continuous norming method enhanced the quality of norms significantly for most contemporary psychological assessment scales.Currently,true score equating or linking method based on Item Response Theory(TRT)shows great advantages in norm development.This paper explored several strategies that use Gorsuch’s continuous norming method,CTT method and IRT method. Solutions and recommendations for fixing the problems in the norm development are given in the conclusions.更多还原