| AIDS is a chronic disease caused by infection with the human immunodeficiency virus.Once infected,the virus replicates in the body continuously and attacks the body’s immune cells,resulting in damage and dysfunction of the body’s immune system.Over time,the body’s immune system loses its ability to defend itself against various pathogens gradually,making it susceptible to various infections and malignancies.Personalized medicine is the provision of precise treatment plans for patients based on their own characteristics.As patient-level data becomes more available,more accessible and of higher quality,survival data can be used to assess treatment effectiveness and predict patient survival time.In clinical studies with time to event as the primary endpoint,the main concern is to find the optimal treatment regime that maximizes the mean survival time of patients.The issue of choosing essential variables becomes challenging when the dimensions of predictors is high.Effective variable selection can enhance prediction precision and the interpretation of decision regulations.In the thesis,a robust estimation framework based on the least squares loss,combined with inverse probability weighting and propensity scores,is used to estimate the optimal treatment regime and variable selection at the same time.The thesis discusses the simulation of the optimal treatment regime for estimating three methods of variable selection under different sample sizes and censoring rates.Results show that the Group Bridge Lasso method outperforms the Adaptive Lasso and Group Lasso methods in terms of variable selection,non-zero value estimation,and decision accuracy when censoring rates are 15%,30%,and 40%,thus improving the precision of determining the optimal treatment regime.Besides,we plot the KaplanMeier survival curve for patients whose received treatment agree with three variable selection method,using AIDS clinical data sets. |
PDF Full Text Request