Characterize mediator-weighting behavior under temporal dependence and weight truncation

Characterize the behavior of mediator weighting in a single interrupted time series with a deterministic exposure switch, particularly its sensitivity to temporal dependence in mediator-outcome confounders and to weight truncation.

Background

The paper develops stabilized mediator weighting for causal mediation analysis with one aggregate time series and a deterministic intervention onset. Because observations are serially dependent and mediator-outcome confounders may have lagged histories, the authors introduce both a single-day mediator weight and a cumulative weight over recent days. Weight truncation is used to control extreme weights, creating a bias-variance trade-off. The paper notes that the behavior of these weighting strategies under varying temporal dependence and truncation choices has not been fully established, making systematic characterization an unresolved methodological problem.