Sparse estimation of large covariance matrices via a nested Lasso penalty

TL;DR

This paper introduces a novel nested Lasso penalty-based method for estimating large covariance matrices with ordered variables, improving sparsity and adaptively selecting bandwidths for better inverse covariance estimation.

Contribution

It develops a new covariance estimator using a nested Lasso penalty on the Cholesky factor, offering more flexibility and better performance than existing methods.

Findings

Outperforms existing covariance estimators in simulations.

Performs well on real data examples.

Performance gap increases with dimension.

Abstract

The paper proposes a new covariance estimator for large covariance matrices when the variables have a natural ordering. Using the Cholesky decomposition of the inverse, we impose a banded structure on the Cholesky factor, and select the bandwidth adaptively for each row of the Cholesky factor, using a novel penalty we call nested Lasso. This structure has more flexibility than regular banding, but, unlike regular Lasso applied to the entries of the Cholesky factor, results in a sparse estimator for the inverse of the covariance matrix. An iterative algorithm for solving the optimization problem is developed. The estimator is compared to a number of other covariance estimators and is shown to do best, both in simulations and on a real data example. Simulations show that the margin by which the estimator outperforms its competitors tends to increase with dimension.