Page 1 of 1
Adding Lag Depended Variable
Posted: Wed Apr 04, 2018 11:33 am
by borninthenorth
Hi! I encounter this theoretical problem while working on my model. After performing Box-Jenkins, I decided to choose I(1) model since it outperformed IMA(1,1), ARI(1,1), ARIMA(1,1,1) in terms of SIC. After performing LM test I can clearly see that in my I(1) model first degree autocorellation is present. To correct that I would add a LDV, however I am not sure whether I should differentiate the LDV or not (my dependent variable is in first order differences). Thank you in advance.
Re: Adding Lag Depended Variable
Posted: Wed Apr 04, 2018 12:35 pm
by startz
The answer would be yes, use the change in your lagged dependent variable. But haven't you already done that in your ARI(1,1) model?
Re: Adding Lag Depended Variable
Posted: Wed Apr 04, 2018 1:14 pm
by borninthenorth
Thank you for your reply. I have done that in my ARI(1,1) model, but since my SIC was lower in I(1) model I decided to work further on with the latter, but then the issue of auto-correlation appeared. I think I will go back to my ARI(1,1), despite the fact that the correlogram does not deliver any significant lags.