An R package VIGoR for joint estimation of multiple linear learners with variational Bayesian inference

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• High-dimensional data and multimodal explanatory variables present challenges for traditional regression models.
• Integrating diverse linear learners into a single framework is computationally intensive.

Purpose of the Study:

• To develop an R package for implementing multiple linear learners within a unified model.
• To facilitate the incorporation of multimodal and high-dimensional variables in regression analysis.

Main Methods:

• Utilizes variational Bayesian inference for model fitting.
• Employs fast minorize-maximization algorithms for efficient computation.
• Implements penalized regression and regression with spike and slab priors.

Main Results:

• The VIGoR package provides a unified approach to complex regression modeling.
• Efficient solutions are obtained even with high-dimensional and multimodal data.
• The package integrates various linear regression techniques.

Conclusions:

• VIGoR offers a powerful tool for advanced regression analysis in R.
• The package simplifies the analysis of complex datasets with numerous predictors.
• It enables more robust and flexible modeling through integrated Bayesian methods.