A procedure, termed modified parallel analysis, is proposed for determining when violations of the assumption of unidimensionality in item response theories are too severe to allow satisfactory parameter estimation. Several Monte Carlo simulation experiments show that the procedure can detect violations of unidimensionality that interfere with parameter estimation. An example of the use of modified parallel analysis is presented in which the comparability of English and French versions of the Otis-Lennon School Abilities Test was examined using data from 1,239 junior and senior high school students. (19 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved).
- modified parallel analysis, detection of violations of unidimensionality that interfere with parameter estimation in item response models
ASJC Scopus subject areas
- Applied Psychology