NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED374158
Record Type: Non-Journal
Publication Date: 1994
Pages: 31
Abstractor: N/A
Reference Count: N/A
ISBN: N/A
ISSN: N/A
Comparing BILOG and LOGIST Estimates for Normal, Truncated Normal, and Beta Ability Distributions.
Abdel-fattah, Abdel-fattah A.
The accuracy of estimation procedures in item response theory was studied using Monte Carlo methods and varying sample size, number of subjects, and distribution of ability parameters for: (1) joint maximum likelihood as implemented in the computer program LOGIST; (2) marginal maximum likelihood; and (3) marginal Bayesian procedures as implemented in the computer program BILOG. Normal ability distributions provided more accurate item parameter estimates for the marginal Bayesian estimation procedure, especially when the number of items and the number of examinees were small. The marginal Bayesian estimation procedure was generally more accurate than the others in estimating a, b, and c parameters. When ability distributions were beta, joint maximum likelihood estimates of the c parameters were the most accurate, or as accurate as the corresponding marginal Bayesian estimates depending on sample size and test length. Guidelines are provided for obtaining accurate estimation for real data. The marginal Bayesian procedure is recommended for short tests and small samples when the ability distribution is normal or truncated normal. Joint maximum likelihood is preferred for large samples when guessing is a concern and the ability distribution is truncated normal. Five tables and 27 figures present analysis results. (Contains 30 references.) (Author/SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A