Estimation in generalized linear models under censored covariates with an application to MIREC data

Area of Science:

• Biostatistics
• Environmental Health
• Clinical Research

Background:

• Biomarker values are often nondetectable (left-censored) due to low concentrations or analytical limitations.
• Accurate analysis of such data is critical in clinical and environmental studies.
• Existing statistical methods may not fully address the challenges of left-censored data.

Purpose of the Study:

• To develop a novel statistical method for maximum likelihood estimation in generalized linear models with left-censored covariates.
• To evaluate the performance of the proposed method against existing estimators through simulations.
• To apply the new method to a real-world environmental health dataset.

Main Methods:

• Development of a maximum likelihood estimation technique for generalized linear models.
• Simulation studies comparing the proposed method with existing estimators.
• Application to a cohort study investigating chemical mixtures and health outcomes.

Main Results:

• The proposed statistical method demonstrates effective estimation for left-censored data.
• Simulations show the new estimators perform favorably compared to existing ones.
• The method provides valuable insights into chemical mixture impacts on maternal and infant health.

Conclusions:

• The novel statistical method offers an improved approach for analyzing biomarker data with detection limits.
• This advancement is significant for environmental health and clinical research.
• The findings contribute to understanding the health effects of chemical exposures.