Login

Publications  •  Project Statistics

Glossary  •  Schools  •  Disciplines
People Search: 
   
Title/Abstract Search: 

Dissertation Information for Akihito Kamata

NAME:
- Akihito Kamata

DEGREE:
- Ph.D.

DISCIPLINE:
- Educational Psychology

SCHOOL:
- Michigan State University (USA) (1998)

ADVISORS:
- None

COMMITTEE MEMBERS:
- None

MPACT Status: Incomplete - Inspected

Title: Some generalizations of the Rasch model: An application of the hierarchical generalized linear model

Abstract: In this dissertation the Rasch model is generalized as a special case of the hierarchical generalized linear model (HGLM), facilitating various extensions. First, the standard binary-response Rasch model is reformulated according to the HGLM specifications. Since the reformulated model is a special case of the logistic HGLM, the model is referred to as a one-parameter hierarchical generalized linear logistic model (1-P HGLLM). Illustrative analyses using hypothetical data sets reveal that parameters estimated via the HLM program are very similar to the estimates from the BILOG program. A parameter recovery study reveals that both item- and person-parameters are estimated properly by the HLM program. Three extensions are presented, including (a) a model with a person-level predictor, (b) a multidimensional model, and (c) a multi-level model. In the first extension, a person-level predictor is added to the model such that the person abilities, as well as the item parameters, are decomposed as a linear combination of more than one parameter. A coefficient of the person-level predictor is properly estimated in the parameter recovery study, as well as in an illustrative analysis. In the second extension, a multidimensional model is formulated. It is shown that the correlation coefficients between latent traits are properly estimated by the HLM program, and that multidimensional analysis can distinguish people who have the same raw scores. In the last extension, a three-level 1-P HGLLM, with an additional level, for schools, is formulated. An illustrative analysis demonstrates that the empirical Bayes estimation enables one to distinguish people who have the same raw scores based on which school each individual has attended. Also, the analysis shows that the empirical Bayes estimation in the three-level model can improve estimation of school means, as well as person abilities. Contributions of this work include (a) a pedagogical presentation of this formulation to the learner of item response theory (IRT) models to facilitate the understanding of IRT models from a different perspective, (b) an easy application of this formulation to conduct a one-step analysis of binary-response test data, (c) a readily accessible multidimensional, as well as multi-level analysis of binary response test data. Several suggestions for future research are also mentioned.

MPACT Scores for Akihito Kamata

A = 0
C = 2
A+C = 2
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2008-01-31 06:05:56

Advisors and Advisees Graph