# Why do I have the same survival probabilities for two individuals having event and no-event using cox mixture model

I have developed a Cox Mixture Model using the smcure() package in R. But the problem here is that when I predict the survival probabilities of each individual using the predictsmcure() in R, a problem arises- I find that the survival probabilities of a person who has experienced an event in say 8th day and one who has not experienced any event upto the end of the observation period and has the same duration till the end of the observation period , i.e 8 days ,have the same survival probabilities. It is as if saying that the survival probabilities are being clustered according to the their time durations upto the censoring time. That means those who have time=5 have the same probabilities, those who have time=10 have the same probabilities and so on, the survival probabilities are being clustered just as in k means clustering according to their survival probabilities.Why does this happen? Given that each individual has different values in the covariates, how can this clustering of survival probabilities be interpreted? Is there any way to obtain the exact survival probabilities of each individual for mixture cure model using coxph in R?

I think what you are doing is obtaining average survival probabilities through predictsmcure() in R. Through this you are actually obtaining the average survival probability on a given time and that remains same across the individuals irrespective of the measures of your covariates as it already takes that into account when you are specifying the covariates in the ~ part of the smcure(). In fact when you are applying the newx= and newz= arguments in the predictsmcure() in R, you are actually specifying the mean or median value of your attributes versus their minimum values like this newx=c(0.8,0) meaning the mean of the attribute which is 0.8 versus value =0 and in case of categorical variables it is like newx=c(1,0) meaning the variable with that factor taken versus not taken. Hence there itself you are taking this average probabilities into consideration.