Testing the ruler with item response theory: Increasing precision of measurement for relationship satisfaction with the Couples Satisfaction Index.

TLDR

The study critically examines the construct of relationship satisfaction in couples research. The authors discuss implications for research. Eight well‑validated self‑report measures of relationship satisfaction, including the MAT and DAS, were administered to 5,315 online participants, and principal‑components analysis and item response theory were applied to a larger item pool to develop the Couples Satisfaction Index (CSI) scales. Item response theory revealed that the MAT and DAS had relatively poor precision, whereas the newly developed CSI scales demonstrated higher precision, stronger convergent validity with other satisfaction measures, and excellent construct validity with anchor scales.

Abstract

The present study took a critical look at a central construct in couples research: relationship satisfaction. Eight well-validated self-report measures of relationship satisfaction, including the Marital Adjustment Test (MAT; H. J. Locke & K. M. Wallace, 1959), the Dyadic Adjustment Scale (DAS; G. B. Spanier, 1976), and an additional 75 potential satisfaction items, were given to 5,315 online participants. Using item response theory, the authors demonstrated that the MAT and DAS provided relatively poor levels of precision in assessing satisfaction, particularly given the length of those scales. Principal-components analysis and item response theory applied to the larger item pool were used to develop the Couples Satisfaction Index (CSI) scales. Compared with the MAS and the DAS, the CSI scales were shown to have higher precision of measurement (less noise) and correspondingly greater power for detecting differences in levels of satisfaction. The CSI scales demonstrated strong convergent validity with other measures of satisfaction and excellent construct validity with anchor scales from the nomological net surrounding satisfaction, suggesting that they assess the same theoretical construct as do prior scales. Implications for research are discussed.