|
Activity Number:
|
290
- Advanced Bayesian Topics (Part 3)
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #318371
|
|
|
Title:
|
A Novel Finite Mixture Model to Cluster Dynamic Latent Ability in Item Response Theory Models
|
|
Author(s):
|
Jingyu Sun* and Xiaojing Wang
|
|
Companies:
|
University of Connecticut, Department of Statistics and University of Connecticut
|
|
Keywords:
|
Item response theory;
Dirichlet process;
Mixture of finite mixture modeling;
MCMC computation;
Dynamic item response modeling;
Nonparametrics methods
|
|
Abstract:
|
The modeling of dichotomous response data in measurement testing is governed by item response theory, also refer to as modern test theory. Traditional item response theory (IRT) models have great limitations in the analysis of longitudinal dichotomous data collected in computerized testing. The dynamic item response (DIR) model proposed by Wang et al.(2013) is a new IRT model that can be applied in such occasions to recover (or predict) ability throughout the time for each person either retrospectively (or on-line). Based on the DIR model, we develop a novel finite mixture model to group the ability trajectory. Our work is the first to introduce clustering ideas for the dynamic changes of latent ability in the study of computerized testing. To test our proposed model, we conduct two different simulation studies to show the accuracy of our methods to cluster the group of individuals according to their latent trajectory. Then, we successfully apply our proposed data to a real dataset from a personalized literacy learning platform called EdSphere. The real data results provide insights for the parents and educational practitioners to make a study plan tailored to students' need
|
Authors who are presenting talks have a * after their name.
