ARIMA Identification

Our goal now is to try to fit a time series data to an appropriate ARIMA process. We use the following approaches to determine a reasonable process to use:

Plot the time series (e.g. as shown in Figure 1 of Stationary Processes): This helps identify trends, which generally require differencing. We generally restrict ourselves to first or second-order differencing.

Calculate ACF and PACF: As we have seen, AR processes have ACF values that converge to zero as the lag increases. MA processes have PACF values that converge to zero as the lag increases. The order of the process may not be obvious using this approach.

AR(p) processes have PACF values that are small (near zero) for lags > p,. MA(q) processes have ACF values that are small for lags > q.

If the ACF and PACF values don’t seem to converge to zero, then differencing may be needed.

If all the ACF values are near zero, then the time series is probably random. We can model such processes as yi = φ0 + εi (white noise process).

When all the ACF values of first differences are near zero, then the time series is probably a random walk, which can be modeled as yi = φ0 + yi-1εi.

6 thoughts on “ARIMA Identification”

  1. Hello sir, I have one problem my data was nonstationary then I was checked the ADF test then I get the third difference becomes stationary but ACF will be decreasing, then this is stationary or not sir and I have also checked the Arima function, I get the second difference becomes stationary. Which one is the best or wrong sir? Please help me sir.

    Reply
  2. Hi Charles, iam trying to use the SARMA_RES and the SARMA_PRED function in Excel but i doesnt seem to work. The other functions such ass ADIFF works.. Do you got any tips? cheers remko

    Reply

Leave a Comment