Regular Learner
Posts: 1

# PROC ARIMA in mathematical form

Hi everyone, I am trying to take the parameters from an ARIMAX model calculated using PROC Arima, and build a forecasting model in excel. I am using variable Y_d (already differenced) as my response variable. And I am using x_d (already differenced) as an explanatory variable. I am using the following code: proc arima data=test; identify var=x_d; estimate p=1 noint; run; identify var=y_d crosscorr=(x_d) ; estimate input= (1 \$ x_d) p=1 q=1; forecast lead=1 out=n; run; quit; According to this http://support.sas.com/documentation/cdl/en/etsug/60372/HTML/default/viewer.htm#etsug_arima_sect012.... (regression with ARMA error), this should measure the following mathematical formula: y_d(t)= constant + B1 x_d(t) + ((1-theta(1)B)/(1-fie(1)B)) *Error(t) As I want to import the parameter in the excel and build a forecasting model in the excel, is it the following model in regression form? y_d(t)=(constant *(1-fie(1)B) +(1-fie(1)B) * B1 x_d(t) +(1-theta(1)B)* Error(t))/(1-fie(1)B) or the final form... y_d(t)=constant_estimate(from SAS)+ fie(1) y_d(t-1)+ B1 x_d(t) - B1*fie(1)*x_d(t-1) + Error (t)-theta(1)* Error(t-1) ? However using the above final form does not produce the correct forecast in Excel using the parameters from PROC ARIMA. If anyone can help me, it would be grealy appreciated. Thanks.
New Contributor
Posts: 3