Shrinkage Estimation of Large-Dimensional Covariance Matrices using Splines

15:40-17:00, Tuesday, April 14, 2026

I-206, Boxue Building

Geert DHAENE is Professor (Emeritus) of Econometrics at KU Leuven. His research lies at the intersection of econometric theory and applied methodology, with a particular emphasis on panel data models, nonlinear estimation, and nuisance parameters. Professor Dhaene's contributions to econometrics are widely recognized. His publications appear in leading journals including Econometrica, The Review of Economic Studies, Journal of Econometrics, Econometric Theory, etc.

The optimal, but infeasible, rotation-equivariant covariance matrix estimator under Frobenius loss shrinks each eigenvalue of the sample covariance matrix individually. We propose a feasible version by approximating the highly nonlinear shrinkage function through penalized splines. The resulting estimator has a simple closed-form solution and is, under very mild assumptions, asymptotically equivalent to its oracle counterpart as both the sample size and dimension grow large. Simulation experiments demonstrate the favorable performance of the proposed estimator compared to other rotation-equivariant estimators, including the linear and nonlinear shrinkage estimators of Ledoit and Wolf (2004) and Ledoit and Wolf (2020).

For more information of the seminar, scan the following QR code(s) to join Tencent QQ group (904 544 292) or WeChat group named "IAER Seminar (4)", please.

QQ Group

WeChat Group 

(QR code is valid until April 20, 2026)