Regression analysis of mixed panel-count data with application to cancer studies

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Recurrent event studies frequently encounter mixed data types (panel-count and panel-binary) due to inconsistent data collection.
• Existing semiparametric methods for mixed data, while unbiased, suffer from computational complexity due to numerous nuisance parameters.

Purpose of the Study:

• To develop a computationally efficient semiparametric approach for analyzing mixed panel-count and panel-binary recurrent event data.
• To simplify the estimation procedure by approximating the nonparametric baseline hazard function.

Main Methods:

• A semiparametric proportional means model was adapted for mixed data types.
• The nonparametric baseline hazard function was approximated to reduce computational complexity.
• Simulation studies were conducted to compare the proposed method with existing approaches.

Main Results:

• The approximated method demonstrated comparable performance to the previous semiparametric likelihood-based method.
• The new approach significantly reduced computation time, offering a much faster estimation procedure.
• The simplified method is implementable in standard statistical software like SAS.

Conclusions:

• Approximating the baseline hazard function provides a computationally efficient alternative for analyzing mixed recurrent event data.
• This simplification reduces the burden of nuisance parameters, enhancing the practicality of the method.
• The proposed approach facilitates the use of advanced statistical techniques in routine data analysis.