Abstract: In order to investigate whether a testlet model can be applied into the performance-based assessment on everyday problem solving in Chinese older adults, 450 older adult participants from 60 to 84 years old from Jiangxi province in China were sampled, including 227 males and 223 females. The goodness-of-fit indexes were tested among the graded response model(GRM), the multidimensional GRM, and two bifactor models which have four and ten specific factors, respectively. The graded response testlet model(GRTM)and GRM were used to analyze the measurement data. The results showed that the bifactor model with ten specific factors was better fitted to the measurement data than other models, and that when items were incorporated into 10 testlets, there were 9 testlets whose effect size variances were beyond 0.50. It was concluded that GRTM can be applied to analyze the measurement data of the performance-based assessment on everyday problem solving in Chinese older adults.

Key words: graded response testlet model, graded response model, local dependence, older adults, everyday problem solving.

摘要： 为考查题组模型在实际应用中的性能及对《基于实际操作的老年人日常问题解决能力测验》的项目质量进行分析,选取了450个60到84岁的老年人的测量数据,进行了模型-数据拟合检验,比较了不同模型的拟合优度,并采用等级反应题组模型（GRTM）对测验项目的参数和题组效应进行了估计。结果显示,包含10个特殊因子和一个公共因子的双因子模型相对更拟合测量数据,将项目组成10个题组后,题组的效应非常显著的题组数达到了9个,GRTM相对于等级反应模型对测验的项目质量作了更好的分析,说明GRTM在测量实践中是可行的,未来测验的编制要加强测量模型的研究。

关键词: 等级反应题组模型, 等级反应模型, 局部相依性, 老年人, 日常问题解决。