I found that in the forecast package in R that I can easily incorporate an exogenous variable Y in my ARIMA model meant to forecast X. While I have a general understanding of the kind of process I need to go through in order to determine the right ARIMA model on X, I'm not sure what I need to understand about Y in order to incorporate it as an exogenous variable.

I was wondering if anybody could help me understand the implications of having this variable in my model and whether there are any tests/checks I can perform to determine whether it's a good idea to have it in my model (other than my intuition and subject matter understanding). Any reading material on this would be helpful. Thanks!


If your error is already very low, you don't have to use xreg, otherwise you may increase the error, but it's worth to try. But if your error is high because your model can't capture the seasonality that you want, try to use an exogenous variable (xreg) for forcing an seasonality.

You can use weekends/holidays dates as an xreg, recurring 0/lowest values is an example of the very thing you want your model to learn. But in my experience, simply try to use the X histogram (binned X) as an xreg, before you try anything. I found that fact after my months of confusion.

The intuition is you can see the histogram/bin as a product of the seasonality itself.


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.