Semi-partitioned Generalized Method of Moments for Longitudinal Data with Lagged and Feedback Covariates

We propose a semi-partitioned Generalized Method of Moments (GMM) framework for analyzing longitudinal data with time-dependent covariates, within a marginal modeling paradigm. This approach addresses limitations of both aggregated and fully partitioned GMM models. Aggregated methods obscure temporal dynamics by assuming constant effects, while fully partitioned approaches offer temporal specificity at the cost of increased model complexity and instability–particularly with moderate sample sizes or deep lag structures. Our method distinguishes immediate from lagged effects by estimating contemporaneous coefficients separately and grouping lagged moment conditions into structured sets, while retaining flexibility in the lag-specific effects. This yields a model that is both statistically efficient and interpretable, capturing essential temporal variation while mitigating variance inflation and convergence challenges associated with full partitioning. The framework accommodates feedback, supports both continuous and binary outcomes, and utilizes the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm for reliable optimization. Through simulations, we demonstrate that the semi-partitioned GMM achieves coverage and competitive efficiency relative to fully partitioned models when the grouped-lag structure approximates the underlying lag pattern. Applications to clinical datasets on knee osteoarthritis and adolescent obesity confirm that the method recovers consistent, interpretable effects and avoids instability associated with finely grained partitioning.

💡 Research Summary

This paper introduces a semi‑partitioned Generalized Method of Moments (GMM) framework designed for longitudinal studies that involve time‑dependent covariates, including lagged effects and feedback mechanisms. Traditional approaches fall into two extremes. Aggregated GMM (e.g., Lai and Small) pools all valid moment conditions for a covariate into a single estimating equation, assuming a constant effect over time. While computationally cheap, this masks temporal heterogeneity and can lead to biased or uninterpretable results when effects truly vary across lags. At the opposite extreme, fully partitioned GMM (e.g., Irimata et al.) assigns a distinct regression coefficient to every lag of every covariate, thereby capturing fine‑grained dynamics but inflating the parameter space proportionally to the number of time points and covariates. In moderate‑sized samples or deep lag structures, this can cause singular information matrices, inflated standard errors, and convergence failures, especially for binary outcomes where event counts per lag become sparse.
The semi‑partitioned GMM strikes a balance by estimating the contemporaneous (lag‑0) effect separately while grouping lagged covariate values into a small number of pre‑specified blocks (e.g., “near” vs. “far” lags, first lag vs. all later lags, or a single block summarizing all lags). Within each block, all lagged observations share a common coefficient, reducing dimensionality while preserving the ability to test hypotheses about temporal patterns. The general model can be written as

 g(μ_it) = β₀ + Σ_j