0

I've a time series, 100 observations. I'm following the the Box-Jenkins method in order to find a model that fit the data. My questions (I'm a bit confused):

  1. if I plot in a graph time vs observation how can I say if the series is stationary just looking the graph?

  2. I can use an AR, MA or ARMA model only if the series is stationary. Why?

  3. If the ACF and the PACF "say" that the process is an AR or MA this implies that the series is stationary? If yes, why?

  4. Can I tell if a series is stationary looking at the ACF and the PACF?

  5. Which are the methods to check if a series is stationary?

Stephan Kolassa
  • 95,027
  • 13
  • 197
  • 357
zar
  • 51
  • 1
  • 8
  • 3
    Two earlier questions that I strongly recommend: [Why does a time series have to be stationary?](https://stats.stackexchange.com/q/19715/1352) and [How to know if a time series is stationary or non-stationary?](https://stats.stackexchange.com/q/27332/1352) – Stephan Kolassa May 08 '17 at 13:33
  • 1
    Not all AR, MA or ARMA models are stationary. It depends on the values of the parameters. The ACF and the PACF are useful guides. Look at some of the posts on this topic. The ones Stephan Kolassa recommends provide a good start. – Michael R. Chernick May 08 '17 at 13:49

0 Answers0