Journal article
The Impact of Three Factors on the Recovery of Item Parameters for the Three-Parameter Logistic Model
Applied measurement in education, Vol.30(3), pp.228-242
07/03/2017
DOI: 10.1080/08957347.2017.1316274
Abstract
This article provides a detailed description of three factors (specification of the ability distribution, numerical integration, and frame of reference for the item parameter estimates) that might affect the item parameter estimation of the three-parameter logistic model, and compares five item calibration methods, which are combinations of the three factors, using a simulation study. The five item calibration methods are Normal-Midpoint-Prior (NMPr), Normal-Hermite-Prior (NHPr), Normal-Midpoint-Posterior (NMPo), Normal-Hermite-Posterior (NHPo), and Empirical-Midpoint-Prior (EMPr). In addition, four item response theory computer programs (BILOG-MG, PARSCALE, flexMIRT, and ICL) are compared in terms of their default specifications and available options of the three factors. The EMPr method recovered item parameters accurately regardless of the shape of the population ability distribution and the number of quadrature points. The NMPr, NHPr, NMPo, and NHPo methods returned item parameter estimates with low bias when abilities followed a standard normal distribution, but tended to either underestimate or overestimate the item parameters when the population ability distribution was skewed. Also, unlike the EMPr method, the performance of these four methods depended on the number of quadrature points.
Details
- Title: Subtitle
- The Impact of Three Factors on the Recovery of Item Parameters for the Three-Parameter Logistic Model
- Creators
- Kyung Yong Kim - Univ Iowa, Coll Educ, 210 Lindquist Ctr South, Iowa City, IA 52242 USAWon-Chan Lee - University of Iowa
- Resource Type
- Journal article
- Publication Details
- Applied measurement in education, Vol.30(3), pp.228-242
- Publisher
- Taylor & Francis
- DOI
- 10.1080/08957347.2017.1316274
- ISSN
- 0895-7347
- eISSN
- 1532-4818
- Number of pages
- 15
- Grant note
- DOI: 10.13039/100005246, name: Institute of Education Sciences, award: R324A100026
- Language
- English
- Date published
- 07/03/2017
- Academic Unit
- Psychological and Quantitative Foundations
- Record Identifier
- 9984371264002771
Metrics
4 Record Views