Ercikan, Kadriye; Burket, George; Julian, Marc; Link, Valerie; Schwarz, Richard;
Abstract: This study investigated the psychometric characteristics of tests containing both multiple-choice (MC) and constructed-response (CR) items and compared scores obtained from five different calibration-scoring configurations. The study is based on a set of tests in which the two item types are combined to measure achievement in the areas of Reading, Language, Mathematics, and Science. These tests were administered as part of a tryout for the Comprehensive Tests of Basic Skills, Fifth Edition. The psychometric properties of the tests indicated that MC and CR items assessed similar enough constructs to allow the creation of a common scale and provide a single set of scores based on responses to both item types. The tests had high classical reliabilities, and the two item types calibrated well together. Local dependency analysis did not indicate multidimensionality. When the scores from MC only and CR only tests were compared, all instances of high mean differences and low correlations were identified as resulting from low CR test reliabilities and test information due to short test length and extreme test difficulty. (Author/SLD)
Title: Calibration and Scoring of Tests with Multiple-Choice
and Constructed-response Item Types.
Author: Ercikan, Kadriye; Burket, George; Julian, Marc;
Link, Valerie; Schwarz, Richard;
Identifier: M031990 Weber, Melba
Publication Date: 1996-04-00
Descriptors: *Constructed Response; *Multiple Choice Tests; Psychometrics; *Reliability; Scores; *Scoring; *Test Construction; Test Items
Identifiers: *Calibration; Comprehensive Tests of Basic Skills