RaSE: A Random Subspace Ensemble Framework for Variable Screening

Ye Tian from Columbia University proposes a new variable screening framework called Random Subspace Ensemble (RaSE). The method is designed to identify predictors that are jointly dependent with the response, even when they have no marginal effect. It includes theoretical guarantees like sure screening property and rank consistency, supported by simulation studies and real-data analysis.

Use Cases

• Benchmarking variable screening algorithms based on the described framework.
• Identifying high-order interaction effects in predictive modeling based on the method's capability.
• Reducing dimensionality in ultra-high dimensional settings using the ensemble subspace evaluation approach.

Strengths

• Framework is supported by theoretical properties including sure screening property and rank consistency.
• Method is shown to be effective through extensive simulation studies and real-data analysis.
• Framework can be combined with any subspace evaluation criterion, leading to an array of screening methods.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Columbia University

Collection Method

Proposed methodological framework, likely accompanied by simulation and real-data analysis.