Re-Estimation of Student Ability in Foreign Languages Using the Rasch Model
Abstract
This study investigates the effectiveness of the Rasch psychometric model in improving predictive validity of French language placement testing. Three multidimensional models, multiple regression, discriminant analysis, and an a priori rational weighting scheme were compared to three unidimensional ability estimation procedures, raw score, Rasch model score (non-fitting test items deleted), and the Rasch model score corrected for five test disturbances (guessing, test start-up anxiety, sloppiness, item content and person interaction, and plodding). These six estimation procedures (predictor variables) were applied to a single French placement test used in assigning students to three language courses of differing levels of instruction. The criterion (predicted) variables were the student scores obtained on all tests during one semester of language instruction.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1988
- Accession Number
- ADA197151
Entities
People
- Philip J. Westfall
Organizations
- Air Force Institute of Technology