Analysis of Contingency Tables Involving Multiple-Response Data.

Autor/inn/en	Carlson, James E.; Spray, Judith A.
Titel	Analysis of Contingency Tables Involving Multiple-Response Data.
Quelle	(1986), (31 Seiten)Verfügbarkeit
Sprache	englisch
Dokumenttyp	gedruckt; Monographie
Schlagwörter	Cutting Scores; Data Analysis; Difficulty Level; Error of Measurement; Graphs; High Schools; Language Tests; Mathematical Models; Mathematics Tests; Measurement Techniques; Monte Carlo Methods; Multivariate Analysis; Regression (Statistics); Research Methodology; Scoring; Tables (Data); Test Items; ACT Assessment + Suchen Sie Ihr Suchwort? Auswertung; Schwierigkeitsgrad; Messfehler; Grafische Darstellung; High school; Oberschule; Language test; Sprachtest; Mathematical model; Mathematisches Modell; Messtechnik; Monte-Carlo-Methode; Multivariate Analyse; Regression; Regressionsanalyse; Research method; Forschungsmethode; Bewertung; Tabelle; Test content; Testaufgabe; Assessment; Eignungsprüfung; Eignungstest; Hochschulzulassung
Abstract	This paper discussed methods currently under study for use with multiple-response data. Besides using Bonferroni inequality methods to control type one error rate over a set of inferences involving multiple response data, a recently proposed methodology of plotting the p-values resulting from multiple significance tests was explored. Proficiency categories were based on ACT assessment standard scores on the Mathematics Usage Test and different numbers of categories were compared. Cutting points were chosen in the score distribution so that equal numbers of scores fell in each proficiency category. In addition to loglinear analyses of 40 contingency tables, a logistic regression model in which the actual proficiency scores were included rather than a gross categorization, was employed. Results indicate that including terms for differential discrimination and differential difficulty results in overfitting for many of the items of the ACT Assessment Program Mathematics Usage Test. To examine the performance of methods discussed in this paper for the case in which it is known that there are no true differences, a sample-splitting technique was discussed. Future research suggestions on methods of analysis of multiple-response frequency data included Monte Carlo methods and the use of finite intersection methods. (PN)
Erfasst von	ERIC (Education Resources Information Center), Washington, DC