University of Notre Dame
Browse

File(s) under embargo

7

month(s)

20

day(s)

until file(s) become available

Sample Size Requirements for Metropolis-Hastings Robbins-Monro Item Parameter Calibration in the Multidimensional Graded Response Model

thesis
posted on 2023-03-20, 00:00 authored by Kenneth McClure

This thesis examines factors impacting item parameter calibration for the multidimensional graded response model in high dimensions (e.g., > 5) using Metropolis-Hastings Robbins-Monro estimation. Motivation for the use of high dimensional item response models and the associated computational difficulties are reviewed followed by an examination of existing work looking at item parameter recovery in the multidimensional graded response model. Research suggests that both test and trait characteristics, including potential non-normality of marginal trait distributions, interfere with efficient parameter recovery. The present study is the first to systematically examine the impact of marginal trait skewness and excess kurtosis on item parameter recovery. Specific focus is placed on identify necessary sample sizes to obtain accurate parameter estimates and the associated computational burden. Findings suggest that skewness and kurtosis impact different parameter estimates. Further, both sample size and computational burdens appear feasible for use in applied psychological research.

History

Date Modified

2023-03-21

Defense Date

2023-02-16

CIP Code

  • 42.2799

Research Director(s)

Ross C. Jacobucci

Committee Members

Ying Cheng Zhiyong Zhang

Degree

  • Master of Arts

Degree Level

  • Master's Thesis

Alternate Identifier

1373628005

OCLC Number

1373628005

Program Name

  • Psychology, Research and Experimental

Usage metrics

    Masters Theses

    Categories

    No categories selected

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC