图形推理任务作答规则的多策略诊断  

Multiple-Strategy Diagnosis of Rules for Solving Figure Reasoning Tasks

作　　者：刘铁川[1] 赵玉[2] 戴海琦[3] 

机构地区：[1]赣南师范大学教育科学学院,赣州341000 [2]赣南医学院心理学系,赣州341000 [3]江西师范大学心理学院,南昌330022

出　　处：《中国临床心理学杂志》2016年第3期459-463,469,共6页Chinese Journal of Clinical Psychology

基　　金：江西省教育科学规划重点课题(13ZD2L005);江西省高校人文社会科学研究青年基金项目(XL1302)资助

摘　　要：目的:使用多策略测量模型从新的视角诊断图形推理任务的规则掌握情况。方法:根据言语分析或视觉空间两种策略对瑞文高级推理测验题目所使用的规则进行编码,从题目参数、模型拟合、内部效度等多个角度比较了基于单策略假设的DINA模型和基于多策略假设的Mix-DINA和MS-DINA模型的诊断结果。结果:Mix-DINA模型对瑞文高级推理测验的拟合程度、题目区分度、内部效度最高;DINA模型最差;MS-DINA模型很少考虑多策略过程,高估了部分规则难度。结论:使用Mix-DINA模型分析图形推理测验更合理,不仅能够诊断个体的规则掌握状态,还可估计其策略使用倾向。Objective： To diagnose rules used in solving figure reasoning tasks by utilizing new multiple strategy psycho- metric models. Methods： Items in Raven Advanced Progressive Matrices（RAPM） were coded from the perspective of Ver- bal-Analytic strategy and Visual-Spatial strategy, and different cognitive process were involved in correspond strategies. Three psychometric models were used to fit RAPM. The first model was DINA model which assumed only one strategy. The other two models were Mix-DINA and MS-DINA, which allowed multiple strategies. Item parameters, goodness of fit and internal validity were compared between these models. Results： Mix-DINA model fit RAPM best and provided the highest items＇ discriminable and internal validity. MS-DINA model rarely diagnose subjects＇ rules from multiple-strategy perspec- tive and overestimate some rules＇ difficulty. Conclusion： Our data fit RAPM with Mix-DINA model best, which provides both individual＇s rule mastery state and strategy use tendency.

关 键 词：图形推理 解题策略 认知诊断 Mix-DINA模型 

分 类 号：R395.1[哲学宗教—心理学]