Computational Instability of Inverse of Spatial Covariance Matrices

• Main Authors: Chien-Hung Chen, 陳建宏
• Other Authors: Wei-Ying Wu
• Format: Others
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/28086310670079062764

碩士 === 國立東華大學 === 應用數學系 === 103 === The inverse of the covariance matrix is applied very often in statistics. For example, Gaussian likelihood function involves the inverse of a covariance matrix. The computation on the inverse of the correlation matrix could be instable when the observations become dense. In this paper, we investigate how the computation instability in the calculation of the inverse of the covariance matrix leads to the bad performance in the estimation problem and which factors influence the instable of the covariance matrix when the covariance matrix follows the Matern model.