Conditional standard error of measurement and personality scale scores: an investigation of classical test theory estimates with four MMPI scales

Abstract

Abstract This study is a preliminary examination of the fit of three classical test theory models of standard error of measurement to selected personality scale (MMPI) score retest data. The three models compared are the conventional standard error of measurement formula, Lord’s (1955: Lord, F. M. (1955). Estimating test reliability. Educational and Psychological Measurement, 15, 325–336) conditional standard error of measurement formula, and Keats’ (1957: Keats, J. A. (1957). Estimation of error variances of test scores. Psychometrika, 22, 29–41) conditional standard error of measurement formula. Absolute mean error differences between test and retest MMPI scale scores were calculated from boot-strapping (with replacement) a sample of MMPI scale retest scores for 271 psychiatric patients. It was found that, on average across 5000 samplings, Keats’ conditional standard error of measurement formula produced less average absolute difference between estimated and actual MMPI absolute test–retest scale score differences than did the conventionally-employed standard error of measurement formula, and less average absolute error difference than Lord’s conditional standard error of measurement. Potential implications for practitioners in establishing cut-scores for patient classification are raised.