摘要
项目反应理论中参数估计程序的实现 ,一直是研究现代测量理论的学者们关注的问题。该研究从理论上探讨了多级模型参数估计的实现途径 ,并模拟了 5批不同数量及分布情形的项目及被试参数 ,生成相应的原始得分矩阵 ,对自编程序及国际流行的相关程序进行了严格的比较校验 ,验证结果证明本程序具有精确、稳定的性能 ,并且发现被试量太少将影响参数估计的精确性及稳定性。
There are many effective computer programs for applying IRT models in the world, such as BILOG, RUMM, MULTILOG, PARSCALE. But these programs are not easy to be manipulated, especially for those who are not very familiar with IRT. These programs run in DOS mode, where commands are typed into a text file and then submitted. For the purpose of being more widely used with IRT, we discussed the realization of parameter estimation techniques for IRT multiple category scoring model, and developed a computer program called GES. GES is developed with object-oriented mode, and so it is very easy to be operated. To calibrate the program, analysis of five Monte Carlo simulated data sets were used in this study. And the results showed that GES is valid, also we found that the stable and validity of parameter estimation would be violated if the amounts of sample are too small.
出处
《心理学报》
CSSCI
CSCD
北大核心
2003年第4期555-558,共4页
Acta Psychologica Sinica
基金
国教育科学"十五"规划重点课题 (批准号 :DBB0 10 5 0 1)
国家自然科学基金 (批准号 :60 2 63 0 0 5 )资助项目。
关键词
项目反应理论
多级记分模型
参数估计
GES
Item response theory,multiple category scoring model,parameter estimation,GES
