Quantifying Resilience as an Outcome: Advancing the Residual Approach with Influence Statistics to Derive More Adequate Thresholds of Resilience

Resilience as an outcome is defined as better-than-expected wellbeing and developmental progress in the context of exposure to significant adversity. This definition has been quantified through a process of residualization, the difference between an individual’s observed outcome score and its expected (predicted) outcome score using statistical modeling. This approach can be biased by the presence of individuals that disproportionately influence the thresholds for sample characterization as resilient even when the majority of the sample population shows only normative patterns of coping under stress. Our goal in this paper is to present methods to identify these “influencers” and to control for their impact during model estimation. This technique decreases the likelihood to characterize a population as resilient when there is little evidence of exceptional performance by most individuals within the sample. The proposed influencer-adjusted residual approach to modeling resilience results in more adequate predicted outcome values and residuals than other statistical approaches. Conceptionally, however, this approach to data analysis cannot resolve the debate over whether the threshold for being characterized as resilient should be based on an entire study sample or on a subsample with influencers extracted.