The Variance Components Analysis in Binary Outcome Model, based on Multilevel Model

• Main Authors: Yun Chen, 陳云
• Other Authors: Chong-Yau Fu
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/73294956323709926502

碩士 === 國立陽明大學 === 公共衛生研究所 === 98 === Multilevel model (MLM) is often applied to the data with hierarchical structure. A hierarchical structure is common in large scale survey data, because social structure is hierarchical inherently, such as area, community, or family. In addition, based on MLM, the variance components can be derived from the corresponding variance of each level. In this study, those MLM related estimation methods, Goldsetein, Rodriguez and Goldman; Goldsetein, Rasbash, Rodriguez and Goldman, will be reviewed. A Sphincter-Preserving Surgery (SPS) investigation using the data base of Taiwan’s National Health Insurance (NHI), 1997-2006, is as the illustration. The main issue of this data is to exam the variability of the SPS proportion in the change context over 10 years. Note that this study partition 10 years time into three time periods (I, II and III) and get more informative result.