| Summary: | 碩士 === 國立臺中教育大學 === 教育測驗統計研究所 === 97 === The purpose of this paper is to evaluate the performances of the different subscale scores estimation methods by using the simulation data in two testing design situations, the single test design and the equated test design with non-equivalent groups with anchor test design (NEAT) and balanced incomplete block (BIB). In the single test design, factors taken into consideration include the following: a number of the subscales, the test lengths of the subscales, the correlation coefficients between the subscales, and the sample sizes. In the equated test design, factors taken into consideration include the following: the ratios of the subscales, the correlation coefficients between the subscales, and the sample sizes.
The major findings of this study are summarized as follows:
1. The estimation errors decrease as the correlation coefficients between the subscales increases; however, the estimation errors are not impacted by the sample sizes.
2. In the single test design, the estimation errors increases as a number of the subscales increase and the estimation errors decrease as the test lengths decrease.
3. In the equated test design, the estimation errors increase as the ratios of the subscales increase and the estimation errors with NEAT and BIB are almost the same.
|
|---|
