1

I have 5 series, four nonstationary series and one stationary series.I tested each pair using Engle-Granger (x1,x2),(x1,x3),(x1,x4); (x2,x3) (x2, x4); (x3,x4) and found cointegration only in x2->x4, then i did johansen (x1,x2,x3,x4) and Trace test indicated 1 cointegrating eqn(s) at the 0.05 level

What should i follow now?I saw this thread- VAR or VECM for a mix of stationary and nonstationary variables , but couldnt understand what should i do in eviews

Henrique couto
  • 21
  • 2
  • Which part did you not understand? (Note that questions of the type "How to do something in EViews?" are generally of topic.) Also, I think your results are contradictory; it is impossible that only one pair is cointegrated yet there is cointegration among all four variables. This can happen in sample but never in population, so some of the tests must be giving wrong results. – Richard Hardy Feb 28 '21 at 13:47
  • Some more relevant threads are [here](https://stats.stackexchange.com/search?q=%22I%280%29%22+%22I%281%29%22+cointegration+cointegrated). – Richard Hardy Feb 28 '21 at 13:53
  • @RichardHardy i guess my case would be like (c) right, but i dont know how to do it. Is this a vec? I put some results, so you can see. https://imgur.com/dNFwPWB – Henrique couto Feb 28 '21 at 14:18
  • You would build such models by hand or by cleverly manipulating a function that does VECM where you would supply appropriate transformations of variables and restrict some of the coefficients to zero. Yeah, the latter technique requires a full answer to be explained in enough detail... – Richard Hardy Feb 28 '21 at 14:30

0 Answers0