Examining the Construct Validity for the Multiple-Content Testing Programs

Yuan H. Li,Leroy J. Tompkins
DOI: https://doi.org/10.1207/s15327574ijt0403_2
2004-01-01
International Journal of Testing
Abstract:The primary objective of this study was to examine the construct validity for the 2 multiple-content testing programs-the multiple-choice Comprehensive Tests of Basic Skills (CTBS/5) together with the performance-based Maryland School Performance Assessment Program (MSPAP)-by evaluating the true-score longitudinal associations among multiple-content scores in 1 school district. Each content area measure in the 2 testing programs was closely examined by the following criterion: The true-score correlation between 2 time-period measures of the same content area is higher than its longitudinal true-score correlations with other content areas. This criterion was achieved in 2 (Reading and Mathematics) of 5 CTBS/5 content subtests, as well as 1 (Language) of 6 MSPAP content subtests. The structural equation modeling has been conducted on a multitrait-multimethod correlation dataset, where the traits of Reading and Mathematics were assessed by MSPAP and the old version of CTBS/4. Although convergent validity existed in these 2 measures, there was little evidence to support discriminant validity in both measures.
What problem does this paper attempt to address?