1

I'm trying to figure out how stationarity can be useful when analysing time series data and I can't find any satisfying explanations online.

I understand that since the mean and variance are constant it helps out a lot for the various models you can apply to the data but at a conceptual level, I have formed an explanation and I would like your opinion on it.

The way I see it, stationarity gives you an overview of the dependence structure of your data, how do the relationships look like. If this structure changes overtime then if you try to forecast your model, you will have a problem since your assumptions are no longer valid. Therefore is you have stationarity, you can be sure that if you find a correlation within the data that you can exploit, then it will not vary with time.

How accurate is this view ?

PS: I know there are a lot of similar questions asked but again I can't find anything that works for me :)

bsaoptima
  • 111
  • Indeed, stationarity implies a (fairly strong) form of structural stability that says the relationship between past and present is like the one between present and future, so that insights regarding the former can be used to make predictions for the latter. As to necessity, this related question may be relevant: https://stats.stackexchange.com/questions/583927/is-stationarity-of-data-necesarry-in-order-to-do-any-statistics/583929#583929 – Christoph Hanck Aug 08 '22 at 14:11
  • Thank your response, helps a lot ! I think I now understand it way better. – bsaoptima Aug 10 '22 at 08:25

0 Answers0