PROC ARIMA with exogenous variable

[lizzy28]

Hi SAS Community,

I am running forecast for retail sales using ARIMA model. There is an input vairable available, retail_day, which is an indicator whether a day is a retail date or not: 1 for a retail date, and 0 for non-retail date. A retail date means that stores are open and that retail sales are happening.

As the retail calendar is planned based on the calendar, the future values are known. For example, 2/2/2016 is the first day of the retail calendar month, and 2/29/2016 is the last day/the 24th day of the retail calendar. The retail sales are low at the beginning of each month and high at the end of each month.Please see the sample data as attached.

Is there anyway to incorporate the input variable, retail_day, in the ARIMA model to capture the retail calendar effect? Moreover, I only have 3 year historical data, which makes it hard to do seasonal adjustment. 

Any suggestions?

Thanks a lot!

[YueLi · Accepted solution]

Have you tried using the input option in PROC ARIMA to include the retail_day in the model? There are some examples documented in http://support.sas.com/documentation/cdl/en/etsug/66840/HTML/default/viewer.htm#etsug_arima_gettings...