A Calibrated Sensitivity Analysis for Weighted Causal Decompositions.

Area of Science:

• Causal Inference
• Health Disparities Research
• Quantitative Psychology

Background:

• Traditional causal inference struggles with interpreting disparities in minority groups (e.g., race, sexual minority status).
• Causal decomposition analysis offers a way to study disparities by examining intervenable exposures.
• Existing methods often rely on uncheckable assumptions about unmeasured confounders.

Purpose of the Study:

• To develop a sensitivity analysis for causal decomposition estimators to address unmeasured confounding.
• To enhance the interpretability of causal effects on disparities.
• To assess the impact of parental support on suicidal ideation disparities in sexual minority youth.

Main Methods:

• Developed a sensitivity analysis framework using the marginal sensitivity model.
• Employed percentile bootstrap for constructing confidence intervals.
• Proposed a two-parameter reformulation for improved interpretability of unmeasured confounders.

Main Results:

• The effect of parental support on disparities in suicidal ideation among sexual minority youth was found to be small.
• The estimated effect was sensitive to potential unmeasured confounding.
• The proposed methods provide a more nuanced understanding of causal effects on disparities.

Conclusions:

• The developed sensitivity analysis framework is crucial for evaluating the robustness of causal decomposition findings.
• Further research is needed to identify effective interventions for reducing suicidal ideation disparities in vulnerable populations.
• The findings highlight the need for careful consideration of unmeasured confounding in health disparities research.