|
Activity Number:
|
132
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #304469 |
|
Title:
|
Nonparametric Bayesian Modeling of Scaled Item Response Data
|
|
Author(s):
|
Kristin Duncan and José S. Fuentes*+
|
|
Companies:
|
San Diego State University and San Diego State University
|
|
Address:
|
5500 Campanile Drive, San Diego, CA, 92182,
|
|
Keywords:
|
IRT ; Bayesian ; Nonparametric ; Testing ; Dirichlet ; Graded Response
|
|
Abstract:
|
Item response theory (IRT) models the relationship between individuals' unobservable latent trait or ability levels and their responses to items on a test or questionnaire. Understanding the characteristics of questionnaire items such as how well they discriminate between low and high trait subjects allows us to construct better tests and to obtain better estimates of subjects' trait values. Nonparametric Bayesian methods have recently been applied to dichotomous item response data by using Dirichlet process priors to model the item characteristic curve. We extend this model to scaled response using an adaptation of the Graded Response Model. This extension is nontrivial because the dichotomous model deals only with monotone item response functions while adding more response options produces item response functions that are non-monotone.
|
