Stochastic Functional Estimates in Longitudinal Models with Interval-Censored Anchoring Events

Area of Science:

• Biostatistics
• Longitudinal Data Analysis

Background:

• Longitudinal study timelines often rely on specific event occurrences.
• Undefined timelines, due to interval-censored event times, hinder traditional analysis.
• A gap exists in methodologies for longitudinal data with imprecise temporal references.

Purpose of the Study:

• To extend standard longitudinal models for situations with interval-censored anchoring events.
• To develop a robust method for analyzing longitudinal data with less well-defined timelines.
• To provide consistent and asymptotically normal parameter estimators.

Main Methods:

• Expressing regression parameter estimators as stochastic functionals of estimated event time distributions.
• Developing a hybrid computational procedure combining Fisher's scoring and the Expectation-Maximization (EM) algorithm for linear mixed-effects models.
• Conducting simulation studies and real data analysis to validate the proposed method.

Main Results:

• The proposed method extends longitudinal models to accommodate interval-censored anchoring events.
• Functional parameter estimates are shown to be consistent and asymptotically normally distributed.
• The hybrid computational procedure effectively estimates model parameters.

Conclusions:

• The developed methodology fills a critical gap in longitudinal data analysis for studies with imprecise timelines.
• The approach offers a reliable way to handle interval-censored event times in longitudinal research.
• This work enhances the analytical capabilities for a broad class of longitudinal models.