An Application of Vertical Equating for development of Kasetsart Basic Academic Skills Test for Person with Special needs

Abstract

The Kasetsart Basic Academic Skill Test (KBAST) is a battery test of basic academic skills which measures an individual student. It composed of four skill subtests: Word Reading, Word Spelling, Sentence Comprehension, and Math Computation. The test was assigned in order
to build a vertical scale to equate each grade’s test form to all test forms for grade 1 through
grade 12. The first form for individuals grades 1–3, the second form for individuals grades 4–6, the third form for individuals grade 7–9, and the fourth form for individuals grade 10–12. There are 3 main steps of applying vertical equating techniques using grade-to-grade growth
model with an internal common-item anchor-test data collection design. First, subtests were
constructed over basic skill content of the curriculum in grade 1 through grade 12. The subtest forms were groups of selected items from item bank to math IRT difficulty specifications as well as content and format specification. Second, a set of common items were selected and embedded in the subtest forms to create a Ө-scale of the subtest that spans a range of grades fitted to the grade-to-grade growth model. Third, data were collected from equating sample students of grade 1 through grade 12 from the target population students in Bangkok and Khongan province. Data from individuals were grouped into four groups according to the test form they took. The third group (grade 7-9) was chosen as the base scale. The set of common items were run for item parameter estimates and used these item parameters with a linear transformation to place the item parameters and ability distributions for the group 2 and 4 examinees on the Ө -scale that was established for group 3, then used the same process for group 1. With concurrent estimation, parameters for all items were estimated in a single computer run of BILOG-MG. In this run, the grade group was identified so those ability distributions were estimated for each grade groups. By this process of vertical equating, the response on the 4 forms of each skill subtest can be identified and interpreted the ability on the same ability scale. The quality of KBAST was shown 1) the content validity by performing of a group of content experts on the items that carefully constructed, 2) the construct validity based on the grade to grade growth model was shown the ability distributions of groups of examinees and significant tests. The results found that There were three subtests were completely matched with the growth model significantly, except the subtest of word reading, 3) The coefficient index (α) of the reliability of The four skill subtests were among ranged 0.680-0.950