0

I understood ARIMA but I have an issue with the seasonal component. Imagine we have daily data with weekly seasonality as shown in the screenshot.

If I fit an ARIMA(3,0,3)(3,0,3)7 and try to predict the point at t(36), then the non-seasonal components will use the red-marked values:

Image without seasonal compoenents

But regarding the seasonal component, will it go back 7 steps and then use 3 values (because P and Q are set to 3) and therefore use all the green marked values? Or is it going to simply use the last 3 points at t-7, t-14, and t-21 marked in orange (ignoring t-28)?:

Image with seasonal compnents

Arne Decker
  • 808
  • 1
  • 3
  • 9
  • Offtopic: @Arne: I see you deleted [this post](https://stackoverflow.com/a/69000105/16343464), it was a good answer, just a tiny mistake. Don't hesitate to restore it, I'd be happy to upvote ;) – mozway Aug 31 '21 at 14:07
  • 1
    Thank you @mozway, I undeleted it :) – Arne Decker Sep 07 '21 at 13:20

1 Answers1

0

I found it our myself, in case someone cares: It is going to use the last 3 points at t-7, t-14, and t-21 marked in orange, ignoring t-28.

Arne Decker
  • 808
  • 1
  • 3
  • 9