Semiparametric profile likelihood estimation for continuous outcomes with excess zeros in a random-threshold...

Area of Science:

• Biostatistics
• Biomedical Data Analysis
• Statistical Modeling

Background:

• Semicontinuous data, common in biomedical research, present analytical challenges.
• Existing methods often use two-part models (logistic for zero probability, parametric for positive values).
• These approaches may lack flexibility in capturing underlying biological processes.

Purpose of the Study:

• To propose a novel semiparametric model for continuous data with a point mass at zero.
• To develop a flexible statistical procedure based on competing risks.
• To evaluate the method's performance via simulation and application.

Main Methods:

• A semiparametric model inspired by competing damage and resistance biological processes.
• Derivation of a closed-form profile likelihood using the retro-hazard function.
• Simulation studies to assess finite sample properties.
• Application to pulmonary capillary hemorrhage data in rats.

Main Results:

• The proposed semiparametric model provides a flexible framework for semicontinuous data.
• The method demonstrates favorable properties in simulation studies.
• The model successfully analyzes pulmonary capillary hemorrhage data.

Conclusions:

• The novel semiparametric approach offers an effective and flexible alternative for analyzing semicontinuous biomedical data.
• The method's biological basis enhances interpretability.
• This technique advances the statistical toolkit for complex biological data analysis.