Controlling false positive selections in high-dimensional regression and causal inference.

Area of Science:

• Statistics
• Machine Learning
• Bioinformatics

Background:

• Controlling false positive selections is critical in data analysis.
• High-dimensional data presents unique challenges for variable selection.

Purpose of the Study:

• To present generic methods for controlling false positives.
• To assign p-values in subsampling and sample splitting frameworks.
• To adapt these methods for causal variable selection using observational data.

Main Methods:

• Subsampling techniques
• Sample splitting approaches
• P-value assignment strategies
• Adaptations for causal inference

Main Results:

• Encouraging results demonstrated in regression analyses.
• Methods show promise for identifying causally relevant variables.
• Subsampling and sample splitting effectively control false positives.

Conclusions:

• Generic subsampling and sample splitting methods offer robust control of false positives.
• These techniques are valuable for high-dimensional settings and causal variable selection.
• Further research is needed to address remaining challenges in observational data analysis.