论文部分内容阅读
Because of various reasons, data missing exists in most medical surveys. Wecould not use the method to analyze such missing data as the way we handle the complete observed data. How to handle these missing data is becoming one of the focuses of the statisticians in recent years.The original method of solving this problem is the rejection of the missing data, it means only modeling and parameter estimation of the observed data. Obviously, this method will result in the losing of information of the model. Many statisticians have suggested various methods to the problem. Rubin (1974, 1976, 1978) and Little (1987)presented the basic approaches of analysis of the incomplete data. Dempster, Laird and Rubin (1977) systematically described the EM algorithm under incomplete situation. These all applied the thought of solving missing data for us.This paper is based on the data of Shanghai Children Depression Survey. The outcome of the model is missing. We intend to use aimed data to reset the data, and model the logistic regression model, present the parameter estimation method. And according to the real background of the case, modify the model, to obtain the influence of the covariance to the outcomes.