How to build a regression model of survival data ?

dada

New Member
#1
I would like to predict the response variable using different predictors. So far I have been working with multiple linear regression. However as my data is censored I though that survival analysis might be worth to consider.
Models which I have obtained so far using multiple linear regression where in the form of: Y=a1X1+a2X2+...anXn. Therefore it was very easy to see factors important for the prediction of variable Y. I was wondering if survival analysis can give me similar output ? Which type of modelling would you suggest me to use (e.g. Cox parametric hazard models) ?
I am not familiar with the logistic regression. Is the knowledge about logistic regression necessary to understand how survival analysis works ?
I would very much appreciate your help.
I would be also very grateful if you could suggest me a good source of information on this topic (regression models for survival data).
 
#2
Hey dada.

If you were to use a GLM model for survival data, I would suggest checking if the model is in the exponential family. Typical distributions for survival data include the exponential and Weibull distributions.

According to this:

http://en.wikipedia.org/wiki/Exponential_family

Both are in. The thing though is that these don't account for censoring effects.

I'm assuming you want to get the survival rate based on X variables. What structure are your X variables? (i.e. what distribution/attributes have you assigned to them?)