Hello,
I simulated an ARMA 1,1 process and added a level shift. How can I get closer to the actual parameter estimates of the model if a shift is included?
Data A (Keep=x_int x t);
Retain e1 0 x1 0 x_int_1 500;
Do t=-100 To 10000;
e=Rannor(1);
x=(0.02+0.8*x1-0.4*e1+e); * create ARMA 1,1;
x_int=x_int_1+x; * integrate;
e1=e;
x1=x;
x_int_1=x_int;
If t>0 Then Output;
End;
Run;
Data A;
Set A;
dummy=IfN(t>=6000 & t<=7500,1,0);
x_int_d=x_int-dummy*200; * create shift;
Run;
ODS Graphics On;
Proc Timeseries Data=A Plot=Series;
Var x_int_d x_int;
Run;
ODS Graphics Off;
Proc Arima Data=A;
Title "Model 1: With shift";
Identify Var=x_int_d (1) CrossCorr=dummy;
Estimate p=1 q=1 Noint Input=dummy Method=ML;
Run;
Proc Arima Data=A;
Title "Model 2: No shift";
Identify Var=x_int (1);
Estimate p=1 q=1 Method=ML;
Run;
Thanks&kind regards