|
Activity Number:
|
37
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #303888 |
|
Title:
|
Detection of Outlying Discrimination Parameters in a Hierarchical IRT Model
|
|
Author(s):
|
Sherwin Toribio*+ and James Albert
|
|
Companies:
|
University of Wisconsin-La Crosse and Bowling Green State University
|
|
Address:
|
1725 State St., La Crosse, WI, 54601,
|
|
Keywords:
|
IRT models ; Bayesian methods ; Bayes Factor ; Hierarchical model ; Outlier detection
|
|
Abstract:
|
Item Response Theory (IRT) models are commonly used in educational and psychological testing to assess the (latent) abilities of examinees and the effectiveness of the test items in measuring this underlying trait. Two commonly used IRT models for dichotomous response are the One-parameter and Two-parameter IRT models. When items are expected to have similar discrimination values, the hierarchical IRT model might be preferred over the Two-parameter model because it has fewer parameters to estimate. However, it is important to verify that there are no items with outlying discrimination parameters. In this talk, two different Bayesian procedures to check for outlying discrimination parameters will be presented. The first one makes use of the Bayes factor, while the second one utilizes a mixture prior density. The effectiveness of these methods will be illustrated using simulated data.
|
