Perils of partialing: Can scholars predict residualized variables’ nomological nets?

Objective: Partialing is a statistical procedure in which the variance shared among two or more constructs is removed, allowing researchers to examine the unique properties of the residualized, partialed, or unique portions of each construct. Although this technique is common, its use has been criticized due to the difficulty faced in interpreting residualized variables, especially when the original constructs were highly correlated. The aim of the present study was to test the degree to which psychological researchers from the fields of clinical, social, and personality psychology are able to estimate the nomological network of partialed variables accurately. Method: Variables with intercorrelations of varying magnitudes (i.e., anxiety and depression; antisocial and borderline personality disorders) were used to test whether experts can estimate partialed variables’ nomological networks vis-à-vis basic personality trait profiles. Results and Conclusions: We found that, overall, experts were poor at predicting residualized correlations. Factors such as the intercorrelations among the variables and the magnitude of change in the variables’ nomological nets following partialing impacted experts’ accuracy. Suggestions regarding use of this questionable measurement practice are discussed.