NIDA模型参数估计的MCMC算法实现及模型性能研究.doc

文档介绍：
NIDA 模型参数估计的 MCMC 算法实现及
模型性能研究#
蔡艳,涂冬波*
5
10
(江西师范大学心理学院,南昌,330022)
摘要:NIDA 模型提出的初衷是为了弥补 DINA 模型的不足,达到更好地区分不同认知状态
的被试的目的。因此本文针对 NIDA 模型,利用 Monte Carlo 模拟方法,采用国际上常用的
MCMC 算法实现了模型的参数估计,并探讨了模型的性能和参数估计的影响因素。结果表
明:(1) 模型具有较高的精度和稳健性;(2) 样本容量和属性数都是 NIDA 模型参数估计的
影响因素,样本容量越大估计精度越高,属性数越大,估计精度越低;(3) 在中或大样本容
量,且测量属性数小于 7 时,模型是可供选用的。
关键词:NIDA 模型;MCMC 算法;影响因素;模型性能
中图分类号:B841
15
The parameter estimation and properties of NIDA model
using MCMC algorithm
CAI Yang, TU Dongbo
(Pshchology school of JaingXi normal nuiversity, NanChang,330022)
20
25
30
35
40
Abstract: Because of its advantages over other models, DINA model is considered as one of the
most popular cognitive diagnosis model applied in practice. But some researchers think there’s
some deficiencies existed in DINA model. One of these is that it only could identify the examinees
as two different groups at item level. Given that, NIDA model was developed which could almost
distinguish each group. According to this wonderful characteristic, this paper tries to realize the
parameters estimation of NIDA model, and investigate its properties.
In this paper, the MCMC algorithm and Monte Carlo method are used. The findings showed that:
(1) The MCMC algorithm is a applicable method. It reflects that the model holds relatively
strong robustness and great precisions of parameter estimation.
(2) The sample size