Development of a Method for Handling Doubly-Censored Data in a Latent Growth Curve Modeling Framework

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Longitudinal data analysis often faces challenges with censored observations.
• Doubly-censoring in latent growth curve models (LGCMs) can lead to biased estimates and flawed inferences.
• Existing methods may not adequately address mixed censoring effects in complex longitudinal structures.

Purpose of the Study:

• To develop the Generalized Tobit estimator (GBIT) for handling doubly-censored longitudinal data.
• To evaluate the performance of GBIT within LGCMs under various censoring conditions.
• To investigate the impact of doubly-censoring on covariate effects and outcomes in LGCMs.

Main Methods:

• Development of the Generalized Tobit estimator (GBIT) as an extension of the Tobit model.
• Application of Monte Carlo simulations to assess GBIT's performance and accuracy.
• Empirical data analysis to demonstrate GBIT's utility in real-world scenarios with doubly-censored data.

Main Results:

• GBIT effectively handles doubly-censoring effects within the LGCM framework.
• The proposed estimator yields unbiased estimates even with substantial censoring.
• Simulations confirmed GBIT's ability to provide reliable results in longitudinal studies.

Conclusions:

• The Generalized Tobit estimator (GBIT) is a robust tool for analyzing doubly-censored longitudinal data.
• GBIT improves the accuracy of latent growth curve models by mitigating censoring bias.
• This method is particularly valuable for research areas where data are frequently subject to mixed censoring.