the practical problem is as follows:

  1. There are several objects, say, 20 volcanos, that erupt at some point of time (event occurs).
  2. Every volcano has it's unique distribution of 'waiting time' until next eruption (event) occurs.
  3. Every volcano has some observations (from 15 to 60) on those waiting times until next event.
  4. There's no other events (like, only eruption can occur, nothing more), and every volcano has, basically, no failures (it should erupt sooner or later).

What is the proper framework or analysis toolkit for those type of data? I considered time-series techniques, because those events take place for every object in ordered manner like so: ...waiting time>event>waiting time>event>waiting time... But time series analysis implies, that every observation should take place in equal-spaced timing. And my volcano #1 erupt, say 19.07.2017 18:30:00 and nothing happens until 01.08.2017 02:46:00. Or survival analysis tools are more proper here? Particularly, I'm interested in predicting those eruptions and I'm trying to find most useful covariates for it.

Is survival analysis suitable for it?

Thank you!

  • $\begingroup$ The answer will likely to depend on the specific questions you want to answer. Maybe you could add some information in this respect? $\endgroup$ – Michael M Aug 21 '17 at 14:13
  • $\begingroup$ My main point of concern is predicting those eruptions. I discovered that for some of the objects previous (lagged) event time can be a predictor for a currect event time, but for some is not. Season of the year also affects the means of survival times. $\endgroup$ – user16 Aug 21 '17 at 19:01

Your Answer

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

Browse other questions tagged or ask your own question.