how to determine parameters for SARIMA(p,d,q,P,D,Q) model using ACF and PACF graphs of the time series.

Ieva

New Member
#1
I have data ranging from years 1995 - 2020. The data is further divided into quarters of the year. The time series look like this:

Screenshot 2021-05-30 at 21.48.27.png


It can be seen that there is an upward trend as well as seasonal one. The period for the seasonal trend seems to be 2 quarters or 8 months. Should q then be 2? And how do I know the Q? I am not sure about the p and P too.

The ACF and PACF graphs are confusing as well.



Screenshot 2021-05-30 at 21.56.03.png



The lags is measured as quarters of the year. So, every data point depends on 15 data points before it. Then that would make q = 15? and Q = 3?

Here is the PACF graph:

Screenshot 2021-05-30 at 22.01.05.png

I don't know how to interpret this one. The pacf shows which particular prior value contributes the value at hand, graph seems to indicate that the value is 5, but that's 1 year and 8 months, from initial time series graph the period of the seasonal trend seem to be equal to two quarters.