Bayesian Analysis of Familial Aggregation

• Main Authors: Huan-Jan Chang, 張晃禎
• Other Authors: Chuhsing Kate Hsiao
• Format: Others
• Language: zh-TW
• Published: 2000
• Online Access: http://ndltd.ncl.edu.tw/handle/13058779530639644046

碩士 === 國立臺灣大學 === 流行病學研究所 === 88 === Familial aggregation can be expressed by correlation coefficient. There are several methods to estimate the correlation, including one-way random effects model, method of maximum likelihood and Pearson’s product-moment correlation. These methods estimate a fixed value of correlation, and does not contain any prior information. My research is using Bayesian analysis to make statistical inference about the correlation based on observations and prior information. A real application, the data of continuous performance test score of schizophrenic and their families are used to test if there exists familial aggregation. I also compare one-way random effects model, method of maximum likelihood and Bayesian analysis with data simulated under different true values of correlation and various familial sizes. The estimation correlation of continuous performance test score in schizophrenic families by one-way random effects model, method of maximum likelihood and Bayesian analysis are all over 0.16. It means that there exists aggregation of gene or environments in schizophrenic families. For simulated familial data, Bayesian analysis has small standard error than the other methods.