Research On Test Equating Of College Entrance Examination In Science Mathematics In Hunan Province

Objectives:Using CTT and IRT to equate the College Entrance Examination in Science Mathematics in Hunan Province and the simulated examination; trying to use a new equating model to research equating in simulated examination and College Entrance Examination; providing theoretic and practical Reference to avoid "The test decides the destiny" in College Entrance Examination.Methods:Using common-item non-equivalent group design. Based on CTT, we choose the Tucker linear method, Levine Observed Score equating method, Levine True Score equating method and Equipercentile equating method. based on IRT, we choose "mixed model"-three-parameter model in choice questions and Samejima graded responses model in essay questions.Results:The correlation coefficient between the anchor items and two examinations are 0.778 and 0.725, There is a significant correlation between anchor test and two examinations. In linear equating method, whenω=0.4811, intercept is-2.7136 and slope is 0.9336 using the Tucker linear method; intercept is-5.0561 and slope is 0.9487 using Levine Observed Score equating method, using ST and Polyst_wcv1.0 to estimateα-andβ-constants for transforming.choosing Stocking-Lord method to estimate constants:α=1.0216,β=0.1008 in choice questions andα=0.6943,β=-0.3151 in essay questions. In linear equating methods, the equating standard error is minimal(1.6425)using Tucker linear method; the equating standard error is maximal(5.2639)using Equipercentile equating method. In IRT, the equating standard error is 1.1508 in choice questions and 4.1408 in essay questions using Observed Score equating method;the equating standard error is 1.1924 in choice questions and 3.3592 in essay questions using True Score equating method.Conclusions:In terms of the equating data and sample in this research, linear equating method is superior to Equipercentile equating method, and the Tucker linear method is superior to Levine Observed Score equating method. Equipercentile equating method is Undesirable Because of the greater equating standard error. In College Entrance Examination in Science Mathematics,different weights have little influence on equating in linear methods. In IRT,Using "mixed model" to analysis Science Mathematics in College Entrance Examination is feasible.