Hi,
I am predicting the orders for several products using ARIMAX. The only regressors I have used are the monthly dummies. I have specified seasonality of 52 weeks in the model. Many of the products have predict = F after running proc hpfengine. However, some products do have a forecast.
Investigating further I increased the order of AR (from 1 to 2) for the products that did not have forecast. I see that the products have a forecast with higher AR terms. Further, I removed the monthly dummies and again ran the hpfengine with AR=1. I was able to get forecasts again. I am not sure what's causing the issue. Here's a snippet of the code
proc hpfdiagnose data=work.arima_try
outest=est_baseline
repository=xyz
criterion=rmse
SEASONALITY=52
SELECTEVENT=ALL
DELAYINPUT=0
DELAYEVENT=0
print=none;
by Product;
Forecast baseline_order;
id mydate interval=week end="06Jan2019"D;
trend diff=none sdiff=none;
arimax estmethod=ml outlier=(detect=maybe maxnum=3 MAXPCT=5 siglevel=0.05) p=(0:1)(0:1) q=(0:0)(0:0) num=(0:0) den=(0:0);
input m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11 m12;
run;
/*******************************************/
proc hpfengine data=work.arima_try
inest=est_baseline
outest=outest_baseline
repository=xyz
outfor=mytest
lead=105
out=_null_
task=select(criterion=rmse)
SEASONALITY=52
print=none;
by Product;
id mydate interval=week horizonstart="06Jan2019"D;
Forecast baseline_order;
input m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11 m12 ;
run;
