1

In most examples I find ARIMA parameters (p, d, q) to be very small, often 1 or 2.

Is that in practice enough (for p and q especially), or higher values are used, like 10 or 100?

I have found a method to automatically estimate these values using AIC as criteria, but not sure for what range of values to specify.

I should emphasize that I'm working on a Python solution, so automatic parameter detection in R is not an option.

Additionally, I'm also interested in same thing with Seasonal ARIMA.

Kobe-Wan Kenobi
  • 2,437
  • 3
  • 20
  • 33

1 Answers1

0

You can read about forecast package for R by Rob J. Hyndman and its auto.arima function. The package is described in the paper Automatic Time Series Forecasting: The forecast Package for R. It uses the following defaults

max.p = 5, max.q = 5, max.P = 2, max.Q = 2, max.order = 5, max.d = 2,
max.D = 1, start.p = 2, start.q = 2, start.P = 1, start.Q = 1

Since the package implements what you are trying to do (and does it very well), it would be worth checking.

Tim
  • 108,699
  • 20
  • 212
  • 390