An information matrix goodness-of-fit test of the multinomial logistic regression model in case-control studies

• Main Authors: Hsiu-Ya Chang, 張琇雅
• Other Authors: Li-Ching Chen
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/92438570960801204647

碩士 === 淡江大學 === 統計學系碩士班 === 103 === For multinomial response in case-control studies, the multinomial logistic regression model is popularly used to infer the relationship between disease and risk factors. After reparameterisation, the assumed the multinomial logistic regression model is equivalent to several two-sample semiparametric models in which the log ratio of case to control density function is linear in data. Based on this finding, the semiparametric maximum likelihood estimator is constructed. In order to detect the goodness-of-fit of the multinomial logistic regression model, this thesis extends the idea of White(1982) and Zhang(2001) to propose an information matrix based goodness-of-fit statistic based on case-control data. A bootstrap procedure is presented to evaluate the p-value of the proposed test. Power and size comparisons are performed through some simulations. Finally, this thesis illustrates the information-matrix-based test by analyzing two real datasets.