Influence Analysis For Interval-Censored Data In Regression Model

In many research disciplines, such as reliability analysis, survival analysis and medical research, data is often incomplete, and there is one important type of incomplete data which is called interval-censored data. When the sample data is interval-censored, one of the most important problems that are always concerned is how to analyze the relations between the random variable and the other variables with the interval-censored data, such as correlation analysis and regression analysis.In this paper, we mainly discuss the linear regression model in which the dependent variable is interval-censored, and the influence on the regression model. One of the important methods to deal with the interval-censored data is to amend the censored data, and then use the classic method in complete data with the amended value instead of the censored data.We first analyze the influence on the regression model by the deviation, which arises from the substitution of censored data for the real value, and then propose an amended Least Squares Estimate method, considering how we can amend the initial value, such as the midpoint of the censored interval, in order to get smaller influence on the regression model with the amended data.Data-censoring can be considered as a kind of perturbation on the basis of Fisher information, and the perturbation influence on the regression model is also analyzed. The simulation shows that the amended Least Squares Estimation method is rather good, and it is also feasible using Fisher information to analyze the perturbation influence on the linear regression model with interval-censored data.