The solution proposals for missing data problems in Cox regression model

Survival analysis, which is used in many areassuch as medicine, biology, engineering and economics, allows modeling of the time until the desired event takes place. The most important feature of survival analysis is the presence of censoreddata in the data set. Censored data arises where the desired event does not ocur when the research is terminated. The most commonly used method of survival analysis is Cox regression analysis, which determines factors affecting survival time. In Cox regression analysis, all observation values should be known as in other statistical methods. The fact that some values of the variable of interest in the data set cannot be observed is called as missing data and it is an important problem for researchers. The aim of this study is to introduce the methods that solve the missing data problem and to evaluate the performance of different missing data mechanisms and to compare them in data sets with different sample size and missing rate. For this purpose, data sets with 5%, 10%, 20% and 30% missing rates and N = 50, 100, 300 units of sample size are derived for each missing data mechanism. Thus, Cox regression analysis was performed by using missing data analysis methods in each of these data sets for different scenarios. Cox regression analysis, which is commonly used in clinical studies, has also been applied in the field of economics. Missing data analysis methods were compared over Cox regression analysis using unemployment data as real data set.