I came across this:

Time to event with no censoring - use survival or normal regression?

which answers the subject (question). However, I am also wondering, is survival analysis not also the only (?) or one of the few modelling techniques that can cope with time dependent features such as age?

I am also wondering, if it could cope with missing values, where missing-ness is time dependent. Example: A process starts. We know certain feature values and can start predicting how long the process will take. As the process progresses more information (feature values) will be known. This means that the accuracy/uncertainty of the prediction could be improved/reduced. Are there modeling techniques that can cope with situations like this?



I just came across this:


which looks interesting in this context.


1 Answer 1


You could have a look in the framework of joint models for longitudinal and survival data. These models link a time-to-event with a longitudinal outcome (i.e., what you called feature values), and can account for missing data in the longitudinal outcome and a time-varying covariate in the survival one. From these models you can also obtain the so-called dynamic predictions for the survival that are updated as extra information is collected for the longitudinal one.

For more information on these models, you could have a look at the Journal of Statistical Software papers for the R packages JM and JMbayes.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.