A multiplicative structural nested mean model for zero-inflated outcomes

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Zero-inflated nonnegative outcomes are prevalent across various fields, including healthcare and business analytics.
• Analyzing such data requires specialized models that can handle the excess zeros and the nonnegative nature of the outcomes.
• Time-varying confounders and sequential treatments add complexity to standard statistical modeling.

Purpose of the Study:

• To propose a novel class of multiplicative structural nested mean models for zero-inflated nonnegative outcomes.
• To develop a robust statistical framework for estimating treatment effects in the presence of time-varying confounders.
• To provide a method that is accurate and computationally efficient for analyzing complex longitudinal data.

Main Methods:

• Developed a doubly robust estimating equation for the proposed structural nested mean models.
• Estimated nuisance functions (propensity scores and conditional outcome means) using parametric or nonparametric approaches.
• Modeled conditional means in two parts: probability of positive outcomes and mean outcome given positivity, to enhance accuracy for zero-inflated data.

Main Results:

• The proposed estimator is shown to be consistent and asymptotically normal.
• The variance of treatment effect estimators can be consistently estimated using the standard sandwich formula.
• The method demonstrated strong empirical performance in simulation studies and a real-world application.

Conclusions:

• The proposed structural nested mean models offer a flexible and robust approach for analyzing zero-inflated nonnegative outcomes.
• The doubly robust estimation strategy effectively handles complex dependencies and time-varying confounders.
• The method provides reliable estimates of treatment effects, applicable to diverse fields with similar data structures.