ProSGPV

ProSGPV performs variable selection using second-generation p-values to integrate estimation uncertainty via confidence intervals for robust selection in high-dimensional and correlated-predictor settings across continuous, binary, count, and time-to-event outcomes.

Key Features:

• Second-Generation P-values: Incorporates estimation uncertainty by using confidence intervals to compute second-generation p-values rather than relying solely on point estimates.
• Performance in High-Dimensional Data: Handles settings where the number of predictors (p) exceeds the number of observations (n) and where explanatory variables are highly correlated, applicable to genomics and other large-scale datasets.
• Versatility Across Outcome Types: Applies to continuous, binary, count, and time-to-event outcomes.
• Efficiency and Speed: Operates without cross-validation or iterative procedures, reducing computational overhead.
• Visualization Tools: Includes visualization routines to assess variable selection results and the contribution of selected variables.

Scientific Applications:

• Genomics and bioinformatics: Variable selection for large-scale genomic datasets with correlated predictors.
• High-dimensional statistical modeling: Selection of predictive variables in p>n and correlated-predictor contexts to improve inference and prediction.
• Multi-type outcome analyses: Variable selection across continuous, binary, count, and time-to-event data analyses.

Methodology:

Computes second-generation p-values from confidence intervals to integrate estimation uncertainty into variable selection and does not require cross-validation or iterative procedures.