Estimation in Proportional Odds Models with measurement error covariate data

• Main Authors: Sen-Chieh Ho, 何森傑
• Other Authors: Shen-Ming Lee
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/40022697451878361568

碩士 === 逢甲大學 === 統計與精算所 === 92 === The aim of this paper is to make inferences of the estimates in proportional odds model with the covariates that are subject to measurement errors. We extend the ideas of refined regression calibration (RRC) in Liang and Liu（1991）and of sufficiency score(SS) estimate and conditional score(CS) estimate in Stefanski and Carroll(1985， 1987) with logistics regression model. We also use naive estimate that substitute the surrogate with measurement errors for the unknown covariates in unbiased estimating function. We use a simulation study to characterize the performance of these methods. According to the results of simulation, naive estimate give rise to serious bias, the good performance property of RRC method will hold for the covariates with normal distribution and both SS and CS perform better when the regression coefficient is large. Besides,the performance of SS and CS is obviously better than that of RRC when covariates follow chi-square distribution.