Literaturnachweis - Detailanzeige
Autor/inn/en | Ankenmann, Robert D.; Stone, Clement A. |
---|---|
Titel | A Monte Carlo Study of Marginal Maximum Likelihood Parameter Estimates for the Graded Model. |
Quelle | (1992), (39 Seiten)
PDF als Volltext |
Sprache | englisch |
Dokumenttyp | gedruckt; online; Monographie |
Schlagwörter | Computer Simulation; Estimation (Mathematics); Item Bias; Mathematical Models; Maximum Likelihood Statistics; Monte Carlo Methods; Sample Size; Scoring; Statistical Distributions; Test Length |
Abstract | Effects of test length, sample size, and assumed ability distribution were investigated in a multiple replication Monte Carlo study under the 1-parameter (1P) and 2-parameter (2P) logistic graded model with five score levels. Accuracy and variability of item parameter and ability estimates were examined. Monte Carlo methods were used to evaluate marginal maximum likelihood estimates that the MULTILOG computer program produced for the 1P and 2P logistic graded models. Test lengths were 5, 10, and 20 items. Sample sizes were 125, 150, and 500 examinees for the 1P model; and 250, 500, and 1,000 examinees for the 2P model. Item bias and root mean squared error indicate that a minimum sample size of 500 examinees is required for accurate and stable estimates of the 2P graded model item parameters. For both 1P and 2P models, ability distribution and calibration sample size are not important factors in the estimation of ability parameters. Results are discussed in light of a study by S. P. Reise and J. Yu (1990). Together, the 2 studies provide a fairly complete picture of factors that may influence the use of 1P or 2P graded models. Eight figures and 5 tables present analysis results, and there is a 21-item list of references. (SLD) |
Erfasst von | ERIC (Education Resources Information Center), Washington, DC |