Covariance Estimation Based on the Spectral Decomposition for Repeated Measures Data

• Main Authors: Li-Min Chou, 周立敏
• Other Authors: 丘政民
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/65444714801550812522

碩士 === 國立中正大學 === 數理統計研究所 === 89 === Liang and Zeger (1986) introduced generalized estimating equations (GEE) approach under the framework of generalized linear models. This approach has been widely applied in analysis of repeated measures data. However, Crowder (1995) pointed out that there are some pitfalls of using the GEE approach under a particular parametric assumption on the correlation structure. In this thesis, we propose a method of covariance estimation based on the spectral decomposition in conjunction with a nonparametric smoothing method in estimation of the variance-covariance structure. We combine the GEE approach with the proposed covariance estimation method to improve the estimation of regression parameters without any parametric assumption on the correlation matrix. The finite sample performance of the proposed method is examined via a simulation study.