Computerized Adaptive Testing Using the One Parameter Logistic Model with Ability-Based Guessing / 能力植基猜測之一參數對數模式的電腦化適性測驗

碩士 / 國立中正大學 / 心理學所 / 96 / Guessing is very common for a person when he (or she) doesn’t know the answer. Birnbaum (1968) developed the three parameter logistic model (denoted as 3PL model) to describe such ‘guessing’ behavior. According to the 3PL model, ‘guessing’ and ‘ability’ are independent. However, based on some modest reasons, they are not always independent. Therefore, San Martin, Del Pino & De Boeck (2006) proposed the one parameter logistic model with ability-based guessing (denoted as 1PL-AG model) to describe the behavior of ability-based guessing.
This study tried to apply the 1PL-AG model in the computerized adaptive testing (denoted as CAT) to investigate its properties and performances, and compared with its natural competitor, 3PL model (Because the discrimination parameters were set at one here, we called it 1PL-G model) and Rasch model, to see the effect of adding guessing parameter and ability-based guessing parameter to the accuracy of estimation. Under the framework of 1PL-AG model CAT, we investigated Barrada, Olea, Ponsoda & Abad’s (2007) modified progressive method combined freeze process for controlling exposure rate and Ra1che, Blais & Magis’s (2007) adaptive Bayesian estimator for better estimation on extreme ability.
It was found that (a) the 1PL-AG model resulted in a worse bias on higher abilities, and it was more efficient for the examinees with moderate to low abilities, (b) under the framework of 1PL-AG model CAT, the modified progressive method combined with the freeze process improved bank usage and exposure rate control remarkably, and (c) when the ability distribution was equal to the prior distribution, the expected a posteriori estimator was better than the adaptive posteriori a posteriori estimator. On the contrary, when they were unequal, the adaptive estimator was the better choice.

Identiferoai:union.ndltd.org:TW/096CCU05071031
Date January 2008
CreatorsChia-Lung Lee, 李佳隆
ContributorsWen-Chung Wang, 王文中
Source SetsNational Digital Library of Theses and Dissertations in Taiwan
Languagezh-TW
Detected LanguageEnglish
Type學位論文 ; thesis
Format74

Page generated in 0.0013 seconds