I am working on 35 observations with three variables, the natural log of the variables are non-stationary at levels but they become stationary at first difference. There exists no cointegration between the natural logs of the variables, when i run a VAR model of their first difference I get to have most lag length criteria such SC, LR at 0 lag interval and AIC is usaully very high at 6. What should i do in this case please?

  • $\begingroup$ What is the relation between the title question and the text in the body? What is your actual question? $\endgroup$ – Richard Hardy Mar 12 '18 at 9:43
  • $\begingroup$ I have used VAR for three variables at levels but i have difficulties in choosing the right lag length. And these three variables were all stationary at first difference, but had no cointegration. What should I do in this case, Is it correct for me to develop VAR when my variables are stationary at first difference but not cointegrated? $\endgroup$ – Lionel Mbanda Mar 12 '18 at 18:23
  • 1
    $\begingroup$ Lag 6 would imply 6*3 slope coefficients per equation, which seems too much for a time series of 35 observations. Since you indicated in a comment in another thread that you would like to investigate causality, here is a famous step-by-step instruction by David Giles on conducting Granger causality tests. Try following it and see what it leads you to. By the way, you may edit your post and/or its title to make the two match and to form a clear description of what situation you are in and what goal(s) you have. $\endgroup$ – Richard Hardy Mar 12 '18 at 18:37
  • $\begingroup$ Thank you, for a 35 sample observation model is it preferable for me to use ARDL method for cointegration or the VAR model. Because there exists no cointegration between the variables with johansen method. Could this result be attributed to my small sample size? $\endgroup$ – Lionel Mbanda Mar 13 '18 at 12:15
  • $\begingroup$ If you follow the link I provided above, you would use a VAR. $\endgroup$ – Richard Hardy Mar 13 '18 at 12:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.