Question

In time series analysis, how do you decide on the lag length p and q in the ARMA(p,q) model?

Answer 1

Once the data are in their stationary form or this seems a reasonable approximation, then the AR, MA and ARMA models all approximate each other.Assuming an AR(s) model were computed, then I would suggest that the next step in identification is to estimate an MA model with s-1 lags in the uncorrelated errors derived from the regression. The parsimonious MA specification might be considered and this might be compared with a more parsimonious AR specification. Then ARMA models might also be analysed.

ARMA model is a tool for understanding and predicting future values in Arma series.

It consists of two parts namely :

1.) The auto regressive part.

2.) The moving average part.

The model is usually referred to as ARMA (p, q) model.

Where :

p = The order of the auto regressive part.

q = The order of the moving average part.