Reliability of circumplex axes

Abstract

We present a confirmatory factor analysis (CFA) procedure for computing the reliability of circumplex axes. The tau-equivalent CFA variance decomposition model estimates five variance components: general factor, axes, scale-specificity, block-specificity, and item-specificity. Only the axes variance component is used for reliability estimation. We apply the model to six circumplex types and 13 instruments assessing interpersonal and motivational constructs-Interpersonal Adjective List (IAL), Interpersonal Adjective Scales (revised; IAS-R), Inventory of Interpersonal Problems (IIP), Impact Messages Inventory (IMI), Circumplex Scales of Interpersonal Values (CSIV), Support Action Scale Circumplex (SAS-C), Interaction Problems With Animals (IPI-A), Team Role Circle (TRC), Competing Values Leadership Instrument (CV-LI), Love Styles, Organizational Culture Assessment Instrument (OCAI), Customer Orientation Circle (COC), and System for Multi-Level Observation of Groups (behavioral adjectives; SYMLOG)-in 17 German-speaking samples (29 subsamples, grouped by self-report, other report, and metaperception assessments). The general factor accounted for a proportion ranging from 1% to 48% of the item variance, the axes component for 2% to 30%; and scale specificity for 1% to 28%, respectively. Reliability estimates varied considerably from.13 to.92. An application of the Nunnally and Bernstein formula proposed by Markey, Markey, and Tinsley overestimated axes reliabilities in cases of large-scale specificities but otherwise works effectively. Contemporary circumplex evaluations such as Tracey's RANDALL are sensitive to the ratio of the axes and scale-specificity components. In contrast, the proposed model isolates both components. © The Author(s) 2013.