I try to test the relationships of housing prices in different cities using VAR. I noted one city has unit root problem in first difference so I have to transform it into second difference to avoid the stationarity problem. Other variables are stationery in first difference. When I run VAR, can i use the one variable with second difference while other variables with first differences. many thanks.
Dear Miloudi
I think, ARDL is not a suitable approach in this case. ARDL is invalid in the case of I(2)
I agree with Miloudi. I(2) variables are nowhere to be found.
Regarding the more general problem, it's not a problem to run a VAR(p) where the variables have different order of integration. After all, remember that you are just estimating an OLS equation by equation. If they are of different order, however, you will not be able to test for cointegration.
What kind of unit root tests did you perform?
I transformed into log. then run phillip perron unit roo test, achieved I(1). But when I
run Augmented dickery Fuller test, one of the housing price is stationery at I(2). Is phillip perron test better, more robust thant ADF? Thanks
Well done to all contributors to this question. I will also advise you take the natural log of the variables and the test for unit root again. If they are stationary at mixture of I(0) and I(1), it is highly recommended to employ ARDL in the analysis, since this method analyses using different lags. Using VAR for such analysis will be tedious to interpret. VAR could be used when the variables are of same order
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size