I am using annual time series data to estimate a VAR model in R. However, when I run the command to find the optimal lag length, I get the following error:
Fehler in VARselect(data, lag.max = 10, type = "const") : NAs in y.
Although there are no NAs in the data frame that I ma using to estimate the model. What does it indicate? how can I resolve this issue? Also, is it necessary to check for stationarity before applying VAR?
Thank you!
Double check the data, as you have some NAs, R wouln't tell you otherwise.
Please look at a uni texbook about basic stuff about VARs, on the stationarity, as it depends what your goals are.
The sum(is.na(data)) instruction returns the total number of NAs in the data. You should not get such an error if the data has no NAs.
Do you get same error when you use VAR function? (VAR(data, type = "const", lag.max = 10, ic = "AIC"))
The reason that you got this error is that there are some missing observations in your data set. To find which variable has missing observations, you can use "summary(data)". If any variable includes NAs, the last row of the summary statistics shows "NA'S: # of missing variables).
Alternatively, you can use "sum(is.na(data))" which indicates the total number of missing observations in your data set.
Regarding your question about stationary, I recommend you do the unit root test. If variables in your model are stationary I(1), their linear combinations might be stationary; in this case, your variables are cointegrated.
2. Then use the Johansen Cointegration test, if your variables are cointegrated, then the Vector Error Correction Model (VECM) is the appropriate model. The VECM contains both levels (long-run) and differences(short-run). In fact, the VECM combines the long-term equilibrium relationship between variables (cointegration) with the short-term dynamic adjustment mechanism reflecting how the variables react when they move out of long-run equilibrium. There should be theoretical support for the cointegrating relationship.
If your variables are not cointegrated, then use a VAR indifference.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running