ISSN 0439-755X
CN 11-1911/B
主办:中国心理学会
   中国科学院心理研究所
出版:科学出版社

心理学报 ›› 2006, Vol. 38 ›› Issue (06): 934-940.

• • 上一篇    下一篇

应用项目反应理论创建图形推理测验题库

肖玮;苗丹民;朱宁宁;张青华   

  1. 四军医大学心理学教研室,西安 710032
  • 收稿日期:2005-04-10 修回日期:1900-01-01 出版日期:2006-11-30 发布日期:2006-11-30
  • 通讯作者: 苗丹民

The Development of the Item Bank of Graphic Deductive Test
Based on Item Response Theory

Xiao Wei,Miao Danmin,Zhu Ningning,Zhang Qinghua   

  1. Department of Psychology, The Fourth Military Medical University, Xi’an 710032, China
  • Received:2005-04-10 Revised:1900-01-01 Published:2006-11-30 Online:2006-11-30
  • Contact: Miao Danmin

摘要: 自编235个图形推理测验题目。采用铆测验等值设计,以72个联合型瑞文测验题目为铆题,对初中到大学各能力层次的1733名男性进行了测验。使用BILOG MG3.0(边际极大似然估计)对实测数据进行了分析,采用Logsitic 3参数模型。剔除数据与模型拟合不好的题目以及信息函数最大值小于0.3的题目,最终建立一个包含181道题目的题库。该题库可以用于淘汰智力较低的应征青年

关键词: 题库建立, 项目反应理论, 项目等值, 图形推理测验

Abstract: With the application of high-tech weapons in the military arena and the changes in the pattern of warfare, the future high-tech local wars require much more of soldier’s psychological qualifications. In order to improve the quality of the Chinese soldiers, it is vitally important and also necessary to add psychological measurement system to the physical examination of the enlisted men. The history of military personnel psychological selection shows constructing such system is historical development trend. The Psychological Selection System (Version 1.0) is soldier’s qualification test. The objective of the test is to eliminate recruited young men with low intelligence. The content includes: Chinese Vocabulary Reasoning Test (CVRT), Number Operation Test (NOT) and Digital Search Test (DST). Version 2 needs to add nonverbal test to measure pattern-recognition and spatial reasoning.
Method
235 graphic deductive items imitating Combined Raven’s Test (CRT) were developed and administered to 1,733 males with different education levels -- junior high school, senior high school, technical secondary school, freshman and sophomore. Using Anchor-Test design, the participants were divided into 5 groups. The 235 items were divided into 4 tests. The CRT was used as test 5 while the 72 CRT items were distributed to the 4 tests as anchor items. The items were calibrated using BILOG-MG3.0 (Marginal maximum likelihood estimation and three-parameter logistic model). The scale of test 5 (CRT) serves as the reference in the calibration. The items were then deleted if their data-model fitness were not good or the maximum information were less than 0.3. The Cronbach’s alpha and information of each item were calculated for testing reliability. Correlation coefficients between ability of subjects and their scholastic performance were used as criterion-related validity.
Results
The item bank with 181 items were established with a maximum information between 0.30 and 1.13. Based on the location of maximum information, 181 items were divided into 4 groups: <-2.5(32 items), ≤-2.5 and ≤-1.5(79 items), -1.5< and ≤1.5(69 items), and >1.5(1 items). This means that the item bank can only be used to estimate the subjects whose ability are low, but it is not suitable for measuring high ability subjects. The cutoff scores for candidate of conscription was confirmed at -1.64 based on 5% elimination rate. The reliability and validity for low ability person are satisfactory.
Conclusions
Subjects’ performance on graphic deductive test was affected by education level and by the characteristics of the item. Item contents were found to be difficult to remember and to describe, which suggests that they are good for personnel selection. The degree of difficulty of the items was affected mainly by the number of the principles being used in reasoning process. The item bank have satisfactory reliability and validity for individuals with low ability, rendering it fit for being used for elimination of low ability individuals in recruitment

Key words: development of item bank, item response theory, item equating, graphic deductive test

中图分类号: