What question did this study set out to answer?

The research aims to improve variable selection and shrinkage through a Bayesian interpretation of Lasso in high-dimensional data.

Bayesian Hyper-Lassos With Non-Convex Penalization

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial demonstrates enhanced variable selection in high-dimensional data, suggesting improved model accuracy.

The research aims to improve variable selection and shrinkage through a Bayesian interpretation of Lasso in high-dimensional data.
Developed a family of hyper-Lasso penalty functions based on generalized prior distributions.
Explored properties including the oracle property and implemented an EM algorithm for inference.
Conducted simulations alongside a chemometrics example to validate the approach.
The proposed method shows significant improvements over established procedures in variable selection accuracy.
Posterior distributions are multi-modal, with effective exploration of parameter space using random starting values.
Demonstrated oracle properties that enhance the reliability of regression coefficient estimates.

Griffin et al. (2011) studied this question.

View Full Paper

Ask AI

Bookmark