Validation of the Ambivalent and Purposeful Engagement – Trait Measure

Abstract

The Ambivalent and Purposeful Engagement-Trait Measure (APE-TM) was developed to assess two ways that people may react to stressful social interactions: ambivalent engagement (AE: counterproductive attempts to avoid thoughts and feelings) and purposeful engagement (PE: effortfully approaching and working through thoughts and feelings). We carried out three studies in undergraduate and clinical populations to (i) test the robustness of previous psychometric findings and (ii) determine whether AE and PE are specific to social anxiety. Across three studies, our results indicate that the APE-TM is a psychometrically valid measure of ambivalent and purposeful engagement. However, PE appeared less specific to social anxiety and may represent a more general tendency toward adaptive coping. In contrast, AE was specific to social anxiety in both undergraduate samples and clinical samples of individuals with a variety of anxiety disorders and depression.

Keywords:

• social anxiety
• resilient coping
• ambivalent engagement
• purposeful engagement
• ambivalent and purposeful engagement – trait measure

Notes

1. In the current paper, convergent validity refers to the measure under study significantly correlating or relating to measures that it is theoretically hypothesized to relate to, whereas divergent validity refers to the test under study not correlating or relating to measures that it is theoretically not hypothesized to relate to.

2. Because of concerns that results might be biased by the amount of missing data in the sample, we also conducted all convergent and discriminant validity analyses in MPlus using maximum likelihood estimation of missing data. All results were substantively identical.

3. We conducted a Kolmogorov-Smirnov test on all APE-TM items to test for univariate normality; all items were statistically significant, indicating that all items were univariate non-normal. Because multivariate normality cannot be obtained in the case of univariate non-normality, and because MLM is an appropriate estimator for multivariate non-normality, MLM was chosen as the estimator for our analyses.

4. We also conducted a binary logistic regression in which AE, PE, and resilient coping predicted diagnosis and results were substantively identical to those obtained in multiple regression.

5. In a variance decomposition model consisting of one substantive factor (PEand resilient coping items combined) and two uncorrelated method factors, fit obtained was excellent (CFI = 0.99, TLI = 0.99, RMSEA = 0.04, SRMR = 0.03), providing further support for the idea that PE and resilient coping may be largely overlapping constructs.