Increasing the reliability of ipsative interpretations in neuropsychology: A comparison of reliable components analysis and other factor analytic methods

• Published in: Journal of the International Neuropsychological Society Vol. 10; no. 4; pp. 578 - 589
• Main Authors: FRAZIER, THOMAS W., YOUNGSTROM, ERIC A., CHELUNE, GORDON J., NAUGLE, RICHARD I., LINEWEAVER, TARA T.
• Format: Journal Article
• Language: English
• Published: New York, USA Cambridge University Press 01.07.2004
• Subjects: Adolescent; Adult; Aged; Aged, 80 and over; Alzheimer's disease; Batteries; Battery interpretation; Confidence Intervals; Data Interpretation, Statistical; Difference scores; Factor Analysis, Statistical; Female; Humans; Ipsative; Male; Memory - physiology; Middle Aged; Neuropsychological Tests - statistics & numerical data; Neuropsychology; Predictive Value of Tests; Problem Solving - physiology; Psychomotor Performance - physiology; Regression Analysis; Reliable components analysis; Reproducibility of Results; Verbal Behavior - physiology; Difference scores; Ipsative; Battery interpretation; Reliable components analysis
• ISSN: 1355-6177; 1469-7661
• DOI: 10.1017/S1355617704104049

Ipsative approaches to neuropsychological assessment typically involve interpreting difference scores between individual test scores. The utility of these methods is limited by the reliability of neuropsychological difference scores and the number of comparisons between scores. The present study evaluated the utility of difference scores using factor analytic methods, including reliable components analysis (RCA), equally weighted composites and individual neuropsychological measures. Data from 1,364 individuals referred for neuropsychological assessment were factor analyzed and the resulting solutions were used to compute composite scores. Reliabilities and confidence intervals were derived for each method. Results indicated that RCA outperformed other factor analytic methods, but produced a slightly different factor structure. Difference scores derived using orthogonal solutions were slightly more reliable than oblique methods, and both were more reliable than those from equally weighted composites and individual measures. Confidence intervals for difference scores were considerably smaller for factor methods relative to those for individual test comparisons, due to the greater reliability of factor based difference scores and the smaller number of comparisons required. These findings suggest that difference scores derived from orthogonal factor solutions, particularly RCA solutions, may improve reliability for clinical assessment purposes. (JINS, 2004, 10, 578–589.)