The 35th Language Testing Research Colloquium
Untitled Document
Title: Item Response Theory Modeling with EQSIRT
Date: Tuesday, July 2, 2013
Instructors: Peter M. Bentler (UCLA) and Eric Wu (Multivariate Software Inc.)
Occasionally a researcher or research group may need to evaluate the dimensionality of a set of language items, evaluate possible biases in an existing instrument, develop their own instrument for measuring an important proficiency, etc. Item Response Theory (IRT) is the universally accepted methodology to accomplish such goals. EQSIRT was developed by the authors of the EQS structural equations program and IRT experts under sponsorship of the US National Cancer Institute to allow a wide variety of IRT methodologies to be accomplished in a coherent, user friendly, and identical way across Windows, Mac, and Linux environments. This workshop introduces the EQSIRT program. Workshop participants will be given a temporary EQSIRT license and are encouraged to bring their notebook computer for hands-on experience. Download and license information will be emailed to participants before the workshop.
The workshop will start with dimensionality assessment of the data. EQSIRT provides two ways to assess dimensionality with correlation based and full information based exploratory factor analysis. Then a wide range of IRT models will be discussed. EQSIRT allows Rasch, 1PL, 2PL, 3PL models for binary data. It also supports the Graded Response Model (GRM), Generalized Partial Credit Model (GPCM), Rating Scale Model (RSM), and Nominal Response Model (NRM). These models are available in multiple-group and multidimensional form (except NRM). We will also test whether the normality assumption for the latent trait distribution is violated, and show how to model with a non-normal trait. We will also show how to plot in EQSIRT and to use R to obtain publication quality plots.
Peter M. Bentler received his Ph.D. in Clinical Psychology from Stanford University, spent a postdoctoral year at the Educational Testing Service, and has been at UCLA for more than four decades. A former Chair of the Department of Psychology, he is now Distinguished Professor of Psychology and Statistics. Bentler has been an elected president of the Society of Multivariate Experimental Psychology (SMEP), the Psychometric Society, the Western Psychological Association, and the Division of Evaluation, Measurement and Statistics of the American Psychological Association. He received SMEP's 2005 Sells Award for Outstanding Career Contributions to Multivariate Experimental Psychology and, in 2007, was the recipient (with Karl Jöreskog) of the American Psychological Association's Distinguished Scientific Contribution Award for the Applications of Psychology.
Eric Wu is one of the founders of Multivariate Software, Inc., the primary developer of the EQSIRT program. He worked in Professor Bentler's research group at UCLA since mid-1980s and was responsible for the development of the EQS structural equations program.
Session 1: 9:00 a.m. - 10:30 a.m.Introduction to EQSIRT. EFA with bifactor and other rotations. CFA.
Session 2: 10:50 a.m. - 12:15 p.m.Binary and polytomous IRT models.
Session 3: 1:30 p.m. - 3:00 p.m.Models with non-normal latent trait. Multidimensional and multiple group
IRT models with constraints.
Session 4: 3:20 p.m. - 5:00 p.m.Hands-on with EQSIRT on your computer.

Untitled Document
Untitled Document