应用项目反应理论创建图形推理测验题库

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

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B ILOG 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。

Introduction 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 soldiers 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-MG 3.0 （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.