Joint modelling of accelerated failure time and longitudinal data– A case study in cirrhosis data

• Main Authors: Tzu-Hsine Chaung, 張滋顯
• Other Authors: Yi-Kuan Tseng
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/4pye4e

碩士 === 國立中央大學 === 統計研究所 === 102 === In survival analysis, the most common semi-parametric survival model is Cox proportional hazards model, if the proportional hazards assumption fail, we can use other hazards model to be an alternative model. The accelerated failure time model is an attractive alternative model. When we describe the relationship between longitudinal data and survival time, there exist some problems of measurement error and incomplete covariate history that result in biased estimation for partial likelihood function. In order to solve these problems, we introduce the joint model approach. The model parameter estimation of joint model cannot be obtained directly through the joint likelihood function because the integral outside the likelihood function. Therefore, we conduct EM (expectation maximization algorithm) algorithm to estimate the parameters to overcome the difficulty. Since the standard deviation is not easy to estimate the parameters directly, we use the bootstrap procedure to estimate the standard deviation. In recent literature, data fitted by Cox model or accelerated failure model may lead to different results. Therefore, we use the accelerated failure time model to fit the cirrhosis data and compare the results with the ones obtained Cox model.