0

I am trying to build SARIMAX model, trying to determine the (p,d,q) & (P,D,Q,s) values from ACF, PACF plots. My series is stationary without any differencing being required (based on ADF test). While I understand from PACF plot that p=2 & q=2 (though ACF is decaying exponentially, I believe it could be due to cumulative effect of lags on each other) should be good enough, I am slightly confused as to how I can arrive at the P,Q values for the seasonal part. My data is daily observations collected for about 4 years. From the below seasonal part of the decomposition, I could see that pattern repeats itself for every 7 days. Here is the ACF, PACF plots for the undifferenced original data.

I am also confused on the 's' value in (P,D,Q,s) whether to consider as 365 (as data is daily observations) or 7 ( as the pattern repeats itself every week).

Appreciate your guidance in helping me identify the P,Q values for the seasonal part of the series.

ACF PACF Plot ACF-PACF-After Differencing by 7 days Seasonal Component

Sai Pavan Kumar
  • 1
  • 1
  • Post your data in a csv file and I will try and help. You should know that if you have daily deterministic effects (probably !) , the acf and pacf have to be computed taking into these effects and possible holiday effects . See https://stats.stackexchange.com/questions/313810/simple-method-of-forecasting-number-of-guests-given-current-and-historical-data/313852#313852 – IrishStat Feb 05 '20 at 11:58
  • not sure how you can do this in Python ...Perhaps others can suggest how hybrid models can be identified and estimated . – IrishStat Feb 05 '20 at 12:14
  • Unfortunately due to organization restrictions I can't post the actual data – Sai Pavan Kumar Feb 05 '20 at 14:45
  • code the data … that will mask it …. – IrishStat Feb 05 '20 at 16:18

0 Answers0