本文提出「廣義隱藏式馬可夫模型(GHMM)」與「核平滑化無參數試題反應理論模式(KN-IRT)」之整合模式,兼具參數型與無參數型兩類試題反應理論模式去蕪存菁之優點。亦即:本文所提出之整合模式,兼具下列三大優點: 一、不受限於受試人數需200以上;既適用於受試人數較多之大型標測驗,亦適用於受試人數較少具信效度之自編測驗。 二、除可分析受試能力、各試題特徵曲線外,兼可分析猜測及未答問題。 三、無試題局部獨立之限制,可進而與「試題順序理論模式(IOT)」及「試題關聯結構分析(IRS)」之整合應用。
The main purpose of this study was to provide a hybrid model by integrating Generalized Hidden Markov Model (GHMM) and Kernel Smoothing Nonparametric IRT (KN-IRT). This model can be used for item analysis of both series correlation and series non correlation cases. Furthermore this model can be connected with Item Ordering Theory (IOT) or Item Relational Structure (IRS) for describing the relation between the ability and the ordering strength from one item to another.