Perils of partialing psychosocial risk and protective factors: Effects on construct validity

Multiple psychosocial factors can influence any given health outcome. To establish independent effects, partialing correlated risk or protective factors (e.g., simultaneous inclusion in regressions) is a common strategy. However, partialing can alter construct validity for even well-established measures. The present studies used the interpersonal circumplex (IPC) and elements of the five-factor model as frameworks to evaluate this issue.

In Study 1, a representative sample of 1,356 U.S. adults completed measures of social anxiety, social support, and interpersonal problems. In Study 2, 300 middle-aged and older couples (total n = 600) completed measures of martial adjustment, depression, anger/hostility, and optimism, as well as IPC-based measures of the spouses' typical behavior during couple interactions and interpersonal style (i.e., trait social behavior), additional measures of marital support and conflict, and facets of the five-factor-model neuroticism domain. The (dis)similarity of associations of unadjusted and partialed variables with the external variables indicative of construct validity was evaluated in structural summary model analyses of IPC correlates and in comparisons of correlations with the additional measures.

The primary effect of partialing was a significant reduction in the magnitude of expected associations with external variables in the nomological network, suggesting weaker representation of intended constructs. In some cases, partialing also produced a shift in the constructs assessed by partialed scores.

Partialing can alter the validity of psychosocial measures. Implications for designing, reporting, and interpreting analyses of psychosocial risk and protective factors are discussed. (PsycInfo Database Record (c) 2026 APA, all rights reserved).