Modeling of the UCLA Loneliness Scale According to the Multidimensional Item Response Theory

Abstract

The authors of the University of California Los Angeles Loneliness Scale (UCLA-LS) have considered its structure to be unidimensional. Nevertheless, it has been repeatedly found that response direction can generate groupings of items. In this study, the multidimensional item response theory was applied to examine the psychometric properties of the Argentine version of the UCLA-LS. The participants were 556 residents of the Metropolitan Area of Buenos Aires, Argentina (55% female) with an average age of 43.8. The model fits of three models were compared: the unidimensional Graded Response Model (GRM), the multidimensional GRM, and the bifactor GRM. The results indicated that the bifactor GRM, which considered a general factor and two specific factors related to response direction, was the most adequate one. Loneliness, which was the general factor, explained 70.38% of the common variance. Direct response direction toward loneliness explained 20.09% of the common variance; in other words, it clearly biases item responses, whereas response direction toward companionship has a much lower impact (9.52%).