Research on Semi-parametric Varying Coefficient Modeling and Parameter Estimation for Longitudinal Data
Considering of advanced lung cancer progression rule, semi-parametric varying coefficient model and logistic semi-parametric model is used.Checking index investigating in different time of NSCLC (None Small Cell Lung Cancer) is lead into as input variable, and tumor size and whether tumor progress is used as response variable, in order to explore the mapping relationship of time-changing index and tumor change, BPLS estimation methods is used to estimate the coefficients of the model.In parameters estimating, B-spline approximation is needed, time factor is lead into model by B-spline, and penalize estimation is used for variable selection, this method can finish parameter estimation and multi variables selection.After that, PWLSE estimation method is used for logistic varying coefficient model of longitudinal data, so tumor size is prognosis via checking index by this time, and tumor progression probability can be calculated through checking index of this time, and universal model and separated model are set up for different patients respectively, the model error can be auto-correction by RBF-HMM method.This can aid clinical doctors to know tumor progress situation, and take measure to control tumor serious changing speed.By experiments, we found, this modeling method is reasonable and effective, and it can forecast tumor size effectively and indicate doctor to prolong survival time of patients by taking corresponding medical measure.
NSCLC Longitudinal data Logistic semi-parametric model Variables selection
Hui-min LI Pu WANG
College of Electronic Information and Control Engineering, Beijing University of Technology, Beijing, PRC
国际会议
昆明
英文
185-192
2014-05-01（万方平台首次上网日期，不代表论文的发表时间）