Page 1 of 1
BoxJenk
Posted: Fri Jul 23, 2010 6:56 am
by ivory4
After using Boxjen to estimate ARIMA(p,1,q), I DISPLAY the equation defined in the model. Notice the constant is different from estimate while the coefficients are the same.
Why?
How does @armadlm transform ARIMA into statespace form? I would like to use drawn parameters for adlm and fdlm to replace those generated by @armadlm(a=adlm,f=fdlm) equation
Re: BoxJen
Posted: Fri Jul 23, 2010 10:11 am
by TomDoan
ivory4 wrote:After using Boxjenk to estimate ARIMA(p,1,q), I DISPLAY the equation defined in the model. Notice the constant is different from estimate while the coefficients are the same. Why?
That's explained in the description of
BOXJENK.
BOXJENK estimates the model in a form where the CONSTANT is the mean of the process, not the intercept in the reduced form needed in the equation form.
ivory4 wrote:How does @armadlm transform ARIMA into statespace form? I would like to use drawn parameters for adlm and fdlm to replace those generated by @armadlm(a=adlm,f=fdlm) equation
That's explained in quite a bit of detail in the comments in the procedure file. I would just reset the coefficients for the ARMA equation (use %eqnsetcoeffs) and apply
ARMADLM to the result; nothing that you would do would be any different.