R package PRIMsrc: Bump Hunting by Patient Rule Induction Method for Survival, Regression and Classification

Area of Science:

• Statistics
• Machine Learning
• Bioinformatics

Background:

• Traditional statistical methods often struggle with complex, high-dimensional data.
• The Patient Rule Induction Method (PRIM) offers a non-parametric approach for identifying subgroups with distinct outcomes.
• A unified implementation for diverse response types (survival, regression, classification) is lacking.

Purpose of the Study:

• To introduce PRIMsrc, a novel R package implementing a non-parametric bump hunting procedure.
• To provide a unified framework for analyzing various outcome variables including censored time-to-event, continuous, and discrete responses.
• To offer robust model validation and cross-validation techniques tailored for decision-rule making.

Main Methods:

• PRIMsrc utilizes a recursive peeling procedure with response-specific criteria and stopping rules.
• Model fitting involves cross-validated selection for controlling model size and complexity.
• Validation employs an objective function based on prediction error and specialized cross-validation techniques.

Main Results:

• PRIMsrc successfully implements bump hunting for survival, regression, and classification tasks.
• The package facilitates cross-validated model selection to optimize covariate inclusion and peeling steps.
• It generates cross-validated endpoint estimates and provides comprehensive visualization and diagnostic tools.

Conclusions:

• PRIMsrc offers a versatile and powerful tool for non-parametric subgroup analysis across different data types.
• The R package enhances statistical modeling by providing unified, validated, and computationally efficient methods.
• Its availability as open-source software promotes broader application in scientific research and data analysis.