Subject: | |
From: | |
Reply To: | |
Date: | Sat, 15 Jul 2006 23:23:51 -0500 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
title: A Hierarchical Bayesian Framework for Item Response Theory Models with Applications in Ideal Point Estimation
authors: Ying Lu, Xiaohui Wang
entrydate: 2006-07-15 16:51:25
keywords: item response theory, testlet
response theory, random and fixed effect models, vote cast data,
roll call analysis
abstract: Ideal point estimation, a variation of item response theory models, has been widely used by political scientists to analyze legislative behaviors. However, many existing ideal point estimation research is based on unrealistic assumptions of independence of different individuals' decisions towards all cases/bills and the independence of one's decisions towards different cases/bills. The violation of such assumptions leads to bias and inefficiency in parameter estimation. More importantly, failing to address these assumptions has hampered the ideal point estimation research from offering intuitive and concise explanations on complex legislative behaviors such as multidimensionality, strategic voting, temporary coalitions. In this paper, we extend one testlet response theory model by Bradlow, Wainer and Wang(1999) to a comprehensive hierarchical Bayesian statistical framework that allows
researchers to model inter-individual and intra-individual correlations through
random effects and/or fixed effects. Through simulations and an analysis of the US Supreme Court vote cast data, we show that the proposed framework holds good promise for tackling many unsettled issues in ideal point estimations. As a companion to this paper, we also offer an easy-to-use R package with C code that implements the methods discussed herein.
http://polmeth.wustl.edu/retrieve.php?id=616
**********************************************************
Political Methodology E-Mail List
Editor: Karen Long Jusko <[log in to unmask]>
**********************************************************
Send messages to [log in to unmask]
To join the list, cancel your subscription, or modify
your subscription settings visit:
http://polmeth.wustl.edu/polmeth.php
**********************************************************
|
|
|