The IRT Equating Study Of National Matriculation English Tests

Equating is a topical issue in testing field for a long time. It is also a necessity for large-scale assessment at home and abroad to keep scientific, fair and consistent. As to the National Matriculation Test, such a large scale assessment, the decentralization of the proposition causes the incomparability of different tests, which hurdles the maintenance of the fairness of NMT.This thesis makes a study of the equating model of three representative NMET papers. The equating is based on item response theory Rasch model and its variant models for objective as well as subjective item types. The equivalent-group design is used to collect the equating data.The procedure of this study consists of five steps:1. to select the most suitable tests for equating; 2. to do the model-data fit check of these three tests by the real data collected from the test administration including tests Shanghai 09,08 and Jiangxi 09; 3. to determine the equating design method, to select the person sample, to administer these tests and to get the data that equating needs; 4. to pre-process the raw data and to put them into the Rasch analysis software to estimate the "person ability" of these three groups by Joint Maximum Likelihood Estimation; 5. to deal with the "person ability measures" and to deduce the linear relationship model between Shanghai08 and Shanghai 09, as well as the model between Shanghai 09 and Jiangxi 09.The study results indicate that:1. The selected matriculation test papers fit the one parameter Rasch model in item response theory; 2. The tests of Shanghai 08 and Shanghai 09 maintain good consistency and the difficulty of Shanghai 08 is higher than Shanghai 09. The equating relation model is: Yb=0.8931 X-0.25850.89 X-0.26; 3. The result of the relation between Shanghai 09 and Jiangxi 09 is that, Jiangxi is slight harder than Shanghai for the person whose "ability measure" is below zero; Jiangxi is easier than Shanghai for the person whose "ability measure" is above or equals to zero. The equating relation model is: Yc=1.3166 X+0.1966 - 1.32 X+ 0.20.The significance of this study includes three aspects:firstly, this study makes an exploratory research on the equating method of the NMET papers; secondly, this study works out the equating relation model for Shanghai 08 and Shanghai 09, as well as the relation model for Shanghai 09 and Jiangxi 09; thirdly, this study provides a valuable reference and an important study basis for the future study of comparison and equating of NMET papers, for the maintenance of the consistency of the NMET, for the further study of the comparison of the "person ability measures" from different provinces or from different years in one province.