Psychometric Properties of an Emotional Adjustment Measure An Application of the Graded Response Model Víctor J. Rubio 1 , David Aguado 2 , Pedro M. Hontangas 3 , and José M. Hernández 1 1 Department of Biological and Health Psychology, Autonoma University of Madrid 2 Institute of Knowledge Engineering, Autonoma University of Madrid 3 Department of Methodology of Behavioral Sciences, University of Valencia, all Spain Abstract. Item response theory (IRT) provides valuable methods for the analysis of the psychometric properties of a psychological measure. However, IRT has been mainly used for assessing achievements and ability rather than personality factors. This paper presents an application of the IRT to a personality measure. Thus, the psychometric properties of a new emotional adjustment measure that consists of a 28-six graded response items is shown. Classical test theory (CTT) analyses as well as IRT analyses are carried out. Samejima’s (1969) graded-response model has been used for estimating item parameters. Results show that the bank of items fulfills model assump- tions and fits the data reasonably well, demonstrating the suitability of the IRT models for the description and use of data originating from personality measures. In this sense, the model fulfills the expectations that IRT has undoubted advantages: (1) The invariance of the estimated parameters, (2) the treatment given to the standard error of measurement, and (3) the possibilities offered for the construction of computerized adaptive tests (CAT). The bank of items shows good reliability. It also shows convergent validity compared to the Eysenck Personality Inventory (EPQ-A; Eysenck & Eysenck, 1975) and the Big Five Questionnaire (BFQ; Caprara, Barbaranelli, & Borgogni, 1993). Keywords: emotional adjustment, item response theory, Samejima’s graded response model, personnel recruitment Emotional Adjustment (EA, also called Neuroticism, emo- tional equilibrium, Emotional Stability) is one of the con- structs that systematically appears to determine “personal- ity structure.” It constitutes a dimension in most of person- ality theories (Cattell & Scheier, 1961; Eysenck, 1947; Guilford, 1959). Recently, the five-factor model (FFM) of personality includes Emotional Stability as one of the “Big Five.” Actually, EA is less conceptually controversial di- mension (Costa & McCrae, 1992; Digman, 1990; Wiggins & Trapnell, 1997). EA relates to whether or not someone has the tendency to feel negative emotions and have irra- tional thoughts as well as to control the impulses when fac- ing a stressful situation. Characteristics of this dimension are moody, touchy, irritable, anxious, unstable, pessimistic, and complaining versus controlled, secure, calm, self-sat- isfied, and cool. The dimension EA is also included in most of the commonly used personality questionnaires, such as the 16PF (Cattell, 1972), EPQ (Eysenck & Eysenck, 1975), and NEO-PI-R (Costa & McCrae, 1992). The Use of IRT in Personality Assessment IRT provides valuable methods for the analysis of the psy- chometric properties of a psychological measure. IRT has several advantages compared to classical test theory (CTT). First, the psychometric information is not sample dependent (Lord, 1980). Second, the effectiveness of a scale can be assessed at every level of the trait being mea- sured. Consequently, IRT provides guidance in the con- struction of tests and allows the designer to tailor the effi- ciency of the instrument for a specific level of the trait. Third, each item is not equally weighted in the estimation of the trait level. Thus, the predicted trait level will be equivalent regardless of the items on which it is based. Fi- nally, psychometric development in IRT in conjunction with advances and use-generalization in computer technol- ogy have made computerized adaptive testing (CAT) fea- sible. Despite those advantages, a general overview of the psychometric literature shows that IRT has been mainly used for assessing achievements and ability rather than per- sonality factors (Ferrando, 1994; Cooke & Michie, 1997; Rouse, Finger, & Butcher, 1999). Two reasons have been pointed out for this (Waller & Reise, 1989). The first one states that IRT is without any doubt a more complex and higher mathematically demanding psychometric theory. The second reason implies the fact that, at this moment, the majority of IRT models entail the assumption of unidimen- sionality, that is to say a single trait or factor explains sub- jects’ performance level in a particular test. Some person- DOI 10.1027/1015-5759.23.1.39 © 2007 Hogrefe & Huber Publishers European Journal of Psychological Assessment 2007; Vol. 23(1):39–46