Non parametric covariance model with circular condition and its application

Abstract

The estimation of covariance matrix has received significant attention since covariance matrix is used in many fields. In this paper, we propose to estimate a non parametric covariance model with circular condition. A circular kernel is applied to estimate the covariance matrix and the corresponding asymptotic property is derived. Simulation data and real climate data are used to illustrate the new model. The experimental results show that for circular-index covariance matrix, the new method with the circular kernel can get more accurate estimation than the existing methods that use the Gaussian kernel.

Keywords:

• Conditional covariance matrix
• circular variable
• kernel estimation
• non parametric model

Additional information

Funding

This work is supported by the National Natural Science Foundation of China (Grant No. 11701069, 72033002).