0
$\begingroup$
z1=cbind(da[,1],da[,5])
dz1=diffM(z1,12)
m1a=sVARMA(z1,order=c(1,1,0),sorder=c(0,1,1),s=12,include.mean=F)
m1b=sVARMA(dz1,order=c(1,0,0),sorder=c(0,0,1),s=12,include.mean=F)

I think that there should be slight difference between m1a and m1b due to different initial values. Technically, there should be no difference using the original series and set d=1 (m1a) and using differenced series and set d=0. However, the results are completely different. It looks like m1b is correct. I don't know why. I had the same problem for ARIMA and GARCH. At the end of the day, we need m1a. Usually, we are interested in predicting the original series, e.g. GDP not first-difference GDP.

$\endgroup$
1
  • 1
    $\begingroup$ What do you think about my answer? If it is clear, you may accept it by clicking on the tick mark to the left, otherwise you may ask for further clarification. This is how Cross Validated works. $\endgroup$ Commented Jun 1, 2018 at 20:10

1 Answer 1

1
$\begingroup$

z1 and dz1 differ by one differencing operation, specifically a seasonal difference. Meanwhile, your model specifications in m1a and m1b differ by two differencing operations, one nonseasonal and one seasonal: you have orders (1,1,0)x(0,1,1) and (1,0,0)x(0,0,1).

To match the two, either difference dz1 once more (non-seasonally) or try (1,1,0)x(0,1,1) and (1,1,0)x(0,0,1).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.