Quantile Regression Analysis With Censored Medical Cost

In survival analysis study, the technology about censored lifetime data has been very mature. However, such an outcome is inadequate to assess the treatment or covariate effects on the disease process. To make more comprehensive evaluation, a few secondary outcomes capturing other features of the disease process are often assessed simultaneously. Typical example is the lifetime medical cost. However, analysis of these secondary outcomes has been a statistical challenge. The main difficulty arises from the incomplete follow-up data.This paper introduces the types and the current research status of the lifetime medical cost data, studies regression analysis based the transformation model, proposes a semiparametric transformation model of medical cost. We respectively discuss the regression residuals following normal distribution and extreme distribution. Based on the ideas of likelihood function, we give maximum likelihood estimation of the parameter and the transformation function. We also conduct the simulation research of all kinds of environment, give the estimates of regression parameter. The estimate of the transformation function curve is also given to verify the effectiveness of the proposed model and the estimation method.Another goal of this paper is establishing a linear quantile regression model of medical cost. When quantile and covariate is a linear relationship, we propose a simple weighted estimation equation. Meanwhile, we complete simulation study under the various censored ratio and quantile regression estimation for different parameters to verify the effectiveness of the proposed model and the estimation method.Finally, we apply the two estimators into MADIT data, obtain valid conclusions, which provide powerful help and reference for clinical diagnosis.