The Application Of Survival Analysis Models Combined With Bayesian Model Averaging In Gene Expression Data

In this dissertation,we discuss the application of Proportional Hazards model and Accelerated Failure Time model combined with Bayesian Model Averaging in gene expression data.In addition,this dissertation compares the results of Bayesian Model Averaging with that of Lasso regression.By simulating gene expression data with different censored ratio,this dissertation analyzes the estimation of parameters,the number of parameters whose estimated value are not zero and the predictive log-score.And we evaluate the result of parameter estimation,variable selection and prediction among different models.Finally,this dissertation analyzes the results of two survival analysis models combined with Bayesian Model Averaging by using sarcomas survival data(SARC).This dissertation shows that Bayesian Model Averaging can achieve the purpose of parameter estimation,variable selection and prediction in survival analysis.In addition,the change of censored ratio has little effect on the fitting results of Proportional Hazards model combined with Bayesian Model Averaging.However,with the increase of censored ratio,the results of variable selection and prediction for Accelerated Failure Time model combined with Bayesian Model Averaging are getting worse.