I'm working on a time series forecasting model using VAR (Vector Autoregression). I have 6 features, out of which 2 features are not stationary. If I apply first-order differencing on those features, they are stationary. Should I apply differencing on entire dataset or should I apply on only those 2 features?
$\begingroup$
$\endgroup$
1
-
2$\begingroup$ Without a good reason, do not difference time series that do not have unit roots. This will end up making matters worse (keyword: overdifferencing). $\endgroup$– Richard HardyCommented Mar 10, 2022 at 19:35
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
It all depends on the actual data you're dealing with, generally speaking using VAR with nonstationary time series will not work. However, you can try to approach the problem with VECM.
This could be useful for you: VAR or VECM for a mix of stationary and nonstationary variables?
-
$\begingroup$ Clearly, the OP is not considering using VAR with nonstationary time series. They are considering using VAR with overdifferenced time series instead. $\endgroup$ Commented Mar 11, 2022 at 6:36
-
$\begingroup$ Thank you, I misunderstood the intention. In such case, overdifferencing stationary time series might lead to poor accuracy, instead of that it's worth checking, if simply differencing the nonstationary series and including them in VAR altogether will work. Running several autocorrelation tests is imperative. If that doesn't work, I would suggest using hybrid models or 2 steps approach. It is unclear to me however how the relationships between the series have been determined (e.g. is there any cointegration?), so it's rather difficult to give a straightforward answer. $\endgroup$– FatafimCommented Mar 11, 2022 at 10:18