In testing for stationarity for regression or VAR, does it matter whether the series is stationary without intercept and trend, or with intercept only or with trend only or in any other combination?
Hi,
It is better to test the stationarity with intercept, intercept and trend and none in order to get reliable conclusions. Alternatively, by looking at the graph, you can visually judge whether there is an intercept or trend bor both, depending on that you can select the appropriate test equation. Moreover, it is better to do several stationarity tests such ADF, DF-GLS, KPSS and PP in order to get robust conclusion on the stationarity of the variables.
Hope this help you.
Kind regards
Thushara
Hi there
For robustness check, you might also want to try a variance ratio test (possibly on the demeaned or detrended process)
Good luck!
Rachidi
You need to check all combinations as possible since distributions for testing are not standard. In other words, if the DGP has a constant and trend, this would affect the distribution of the test when you perform a test assuming only a trend in the model. There are packages for doing that easily.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression