MDCcure: An R package for martingale difference correlation and hypothesis testing in mixture cure models.

Area of Science:

• Biostatistics
• Survival Analysis
• Computational Statistics

Background:

• Assessing covariate-clinical outcome relationships is crucial in biostatistics, especially for survival and cure models.
• Traditional methods struggle with complex dependencies and nonparametric evaluation of covariate effects.
• There's a need for efficient computational tools for hypothesis testing and diagnostics in cure models.

Purpose of the Study:

• To present a comprehensive R package for advanced biostatistical analysis of cure models.
• To implement novel and existing methods for dependency analysis, nonparametric hypothesis testing, and goodness-of-fit testing.
• To enhance the estimation of long-term survival and cure probabilities.

Main Methods:

• Developed an R package with functions for martingale difference correlation and divergence to analyze covariate effects on conditional means.
• Proposed a nonparametric framework with four approaches (three martingale-based, one L2 distance-based) for testing covariate significance on cure probability.
• Included goodness-of-fit testing (goft()) and visual comparison tools (plotCure()) for cure rate models.

Main Results:

• The R package demonstrated strong performance in simulations and real-world data analysis.
• Martingale-based measures effectively identified covariates impacting conditional means.
• Nonparametric tests accurately detected covariate effects on cure probability, with enhanced interpretability in multivariable settings.

Conclusions:

• The R package provides accurate and efficient inference tools for cure model analysis.
• The implemented methods improve the assessment of covariate relationships and model fit.
• This facilitates better understanding and estimation of long-term survival and cure probabilities.