rFSA: An R Package for Finding Best Subsets and Interactions.

Area of Science:

• * Computational Biology
• * Biostatistics
• * Data Science

Background:

• * The increasing volume of data in fields like bioinformatics and healthcare challenges traditional model exploration.
• * Existing methods struggle to efficiently explore large datasets for optimal variable selection and higher-order interactions.
• * There is a need for advanced computational tools to facilitate model discovery and improvement.

Purpose of the Study:

• * To introduce the rFSA R package, designed for enhanced variable selection.
• * To present an algorithm that identifies statistically optimal models based on user-defined criteria.
• * To provide a flexible tool for formulating and improving models in data-rich research areas.

Main Methods:

• * Implementation of an iterative algorithm within the rFSA R package.
• * Searching a data space for statistically optimal models of a user-specified form.
• * Evaluation of feasible solutions (candidate models) using quality measures like Allen's PRESS and AIC.

Main Results:

• * The rFSA package provides a set of feasible solutions for researchers to evaluate.
• * The algorithm facilitates the exploration of larger data subsets and higher-order interactions.
• * The package supports linear and generalized linear models, with adaptability for new strategies.

Conclusions:

• * rFSA offers a powerful computational approach for variable selection and model optimization.
• * The package addresses the challenges posed by big data in scientific research.
• * rFSA enhances the ability of researchers to formulate and refine models in diverse fields.