MULTIDIMENSIONAL ITEM RESPONSE THEORY: AN INVESTIGATION OF INTERACTION EFFECTS BETWEEN FACTORS ON ITEM PARAMETER RECOVERY USING MARKOV CHAIN MONTE CARLO

作者: Jonghwan Lee

DOI:

关键词:

摘要: MULTIDIMENSIONAL ITEM RESPONSE THEORY: AN INVESTIGATION OF INTERACTION EFFECTS BETWEEN FACTORS ON PARAMETER RECOVERY USING MARKOV CHAIN MONTE CARLO By Jonghwan Lee It has been more than 50 years since Lord (1952) published “A Theory of Test Scores (Psychometric Monograph No.7)” which is recognized as one the most influential in Item Response (IRT) history. Since then, there extensive research investigating several aspects IRT such as: (1) Modeling; (2) Estimation latent traits; and (3) item parameters. There also development applications based on Equating; Linking; Differential Function (DIF); (4) Standard setting; others. All those have same assumption—that parameters are calibrated accurately possible. Nevertheless, techniques to estimate trait previously developed estimation uni-dimensional model. However, procedures become sophisticated because appearance multidimensional response theory models (MIRT). In MIRT, factors that calibration procedures, number correlation between non-normal distribution different types configurations traits (approximate simple structure mixed structure). this study, interaction effects combined parameter recovery were investigated using Markov Chain Monte Carlo simulation method. The findings show a higher dimensions require bigger sample size lower dimensions—2000 1000 sizes for 6-dimensions 3-dimensions, respectively. That model does not consider skewness distribution, however. This study shows if an additional factor introduced into features or skewness, increasing helpful improving accuracy recovery. Rather, alternative MIRT should be considered case correlated traits, transforming distributions normal distributions. a-parameters affected when traits. d-parameters influence skewed. Overall, model, bias found calibration. If structures independent normally distributed, then dimension specification, less it will true configuration, AS MS, may possibly decrease created from configuration. When suspected having skewed improved by simply size, though might increase items at time. Another way fix problem use examinees selected wide range abilities. with each other. Selecting examinee group carefully greatly reduces resulting procedure. Copyright JONGHWAN LEE 2012

参考文章(64)
Robert L. McKinley, Mark D. Reckase, Some Latent Trait Theory in a Multidimensional Latent Space. ,(1982)
Marek Libura, Multidimensional Item Response Theory Estimation: A computer program Defense Technical Information Center. ,(1987) , 10.21236/ADA197160
Harry Horace Harman, Karl J. Holzinger, Factor analysis : a synthesis of factorial methods University of Chicago Press. ,(1941)
Yanyan Sheng, A MATLAB Package for Markov Chain Monte Carlo with a Multi-Unidimensional IRT Model Journal of Statistical Software. ,vol. 28, pp. 1- 20 ,(2008) , 10.18637/JSS.V028.I10
Mark D. Reckase, 18 Multidimensional Item Response Theory Handbook of Statistics. ,vol. 26, pp. 607- 642 ,(2006) , 10.1016/S0169-7161(06)26018-8
A. Birnbaum, Some latent train models and their use in inferring an examinee's ability Statistical Theories of Mental Test Scores. pp. 395- 479 ,(1968)
Adrian E. Raftery, Steven Lewis, How Many Iterations in the Gibbs Sampler Defense Technical Information Center. ,(1991) , 10.21236/ADA640705
Mark D. Reckase, Robert L. McKinley, The Discriminating Power of Items That Measure More than One Dimension. Applied Psychological Measurement. ,vol. 15, pp. 361- 373 ,(1991) , 10.1177/014662169101500407