| Summary: | 碩士 === 國立臺中教育大學 === 教育測驗統計研究所 === 97 === Testlet items are widely used in the large-scale standardized achievement tests to evaluate the learning achievements of the students. Generally, the item response theories are used to establish the common scales in these large-scale tests; nevertheless, using item response theories to analyze the testlet data will overestimate the precision of measures obtained from testlets and yield biased estimation for item parameters.
The purpose of this paper is to use to evaluate the linking performances of balanced incomplete block design (BIB) and non-equivalent groups with anther test design (NEAT) for horizontal equating designs based on the testlet model by using the simulation data. The factors taken into consideration include the following: the sample sizes, the number of the testlets, the variances of the testlet effects, and the ratios of anchor items.
The results of simulation study show that:
1.The root mean square error (RMSE) of the item parameters decrease as the sample sizes increase.
2.The RMSE of the ability parameters decrease as the number of the testlets increase.
3.The RMSE of the ability parameters increase as the variances of testlet effects increase.
4.The RMSE of the item parameters decrease as the ratios of the anchor items increase.
5.The NEAT design outperforms the BIB design in estimating the item discrimination parameters and the ability parameters.
|
|---|
