عنوان

Comparing three estimation methods for the three-parameter logistic IRT model

Number

TL48643

.Language of Text, Soundtrack etc

انگلیسی

Title Proper

Comparing three estimation methods for the three-parameter logistic IRT model

General Material Designation

[Thesis]

First Statement of Responsibility

Sunil Lamsal

Subsequent Statement of Responsibility

Sheng, Yanyan

Name of Publisher, Distributor, etc.

Southern Illinois University at Carbondale

Date of Publication, Distribution, etc.

2015

Specific Material Designation and Extent of Item

130

Text of Note

Committee members: Asner-Self, Kimberly; Bhattacharya, Bhaskar; Headrick, Todd C.; Koran, Jennifer

Text of Note

Place of publication: United States, Ann Arbor; ISBN=978-1-339-47771-8

Dissertation or thesis details and type of degree

Ph.D.

Discipline of degree

Quantitative Methods

Body granting the degree

Southern Illinois University at Carbondale

Text preceding or following the note

2015

Text of Note

Different estimation procedures have been developed for the unidimensional three?parameter item response theory (IRT) model. These techniques include the marginal maximum likelihood estimation, the fully Bayesian estimation using Markov chain Monte Carlo simulation techniques, and the Metropolis-Hastings Robbin-Monro estimation. With each technique, a prior can be specified to reflect prior belief on each model parameter. Previous studies evaluating the fully Bayesian estimation procedure for this model suggests that the model can be viewed as a mixture model, and it suffers from a non-convergence problem unless strong informative priors are specified for the item slope and intercept parameters. This study focused on comparing the three estimation methods for the three-parameter logistic (3PL) model in parameter estimation using Monte Carlo simulations. In particular, sample sizes, test lengths, prior specifications and actual item parameters were manipulated to reflect various test situations. The results suggest that: 1) all three estimation methods performed differently under different simulation conditions, 2) the three methods performed better when the actual parameters are close to their prior mean or mode than when the actual parameters are close to their boundary values, 3) a relatively more informative prior had to be specified for item parameters to ensure convergence with each of the three methods and the item parameters were more accurately estimated.

Educational tests & measurements

Subject Term

Education;Educational testing;Item response theory

Arab, Pezhman

Sheng, Yanyan

Subdivision

Quantitative Methods

Southern Illinois University at Carbondale

Call Number

1766575451; 10012825

Electronic name

p

[Thesis]

276903

a

Y