Title: High-dimensional Correlation Matrix Estimation for General Continuous Data With Bagging Technique
Time：13 April 2019,16:30
Abstract: High-dimensional covriance matrix estimation plays a central role in multivariate statistical analysis. It is well-known that the sample covariance matrix is singular when p>n, while the covariance estimator must be positive-definite. This motivates some modifications of the sample estimator when considering to remain its nice properties on each entry. In this paper, we modify the sample correlation matrix by using Bagging (Bootstrap Aggregating) technique. The proposed estimator inherits nice properties, e.g., unbiased and efficient, of the sample estimator, and it is flexible for general continuous data. Under few prior assumptions, we show that the estimator can remain positive-definite with high probability in finite samples theoretically. Simulation results and a real application are implemented to demonstrate our method are competitive with other estimators.