On the comparison of models for paired Bernoulli data with extra zeros

• Main Authors: Yu-Hui Chou, 周育慧
• Other Authors: Mi-Chia Ma
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/23457172849595716782

碩士 === 國立成功大學 === 統計學系碩博士班 === 92 === 　　For frequency counts, the situation of extra zeros often occurs in biomedical data. This is demonstrated with count data from a surgical treatment of refractive error which determined the risk factors for an epithelial defect during laser in situ keratomileusis (LASIK). The response variable of this data is binary. It was found that the data exhibited an excess of zeros, in the context that the majority of patients did not suffer intraoperative epithelial defect. The development of surgical treatment is very fast at the time being, count data with extra zeros are quite typical. We use a zero-inflated binomial model to analyze such binary data with extra zeros. In addition, the logistic model, generalized estimating equations, multicategorical logit model and proportional odds model are utilized to analyze this real data. Next, the advantage and defect of different models are compared by the real data and the usage opportunity of above statistical methods is discussed.