5

I'm trying to model with BSTS, with exogenous variables.

It appears that my y-sequence is not stationary. Do I need to diff that to stationary? Or could BSTS handle that?

P.S. I'm using R package bsts

Thank you for your time!

All the best,

Kathy

Chenying Gao
  • 61
  • 4

1 Answers1

9

No, you don't need to make the time series stationary. BSTS should be able to handle that. Stationarity is a requirement specifically for AR and ARMA models.

BSTS is supposed to handle structural changes in the time series, which means that by definition it should be able to handle non-stationary data - since a structural change in the time series would imply changes to the mean and variance of the series.

Skander H.
  • 10,602
  • 2
  • 33
  • 81