0

I have the same problem as the following post, but I have more samples and the index of the outlier is known.

Outlier treatment in Vector Autoregression (VAR) Model

I tried deleting the outliers; it works. I also want to try including dummy variables for the outliers (the dummy take the value of 1 on the outliers, and zero otherwise) and compare the two treatments. One reason I want to do this is that a special event at t1 may also impact the response variable y2 at t2. However, its impact on t2 is not significant enough, thus y2 is not an outlier.

Question: The VAR in R package vars returns "NAs in dummy variable". I guess is because the dummy variable comprises a lot of zeros. I don't know how to fix it. Thanks in advance.

Community
  • 1
user2502240
  • 31
  • 1
  • 3
  • If this is a programing question you might be better off at Stack Overflow. – Andy Jul 15 '14 at 18:26
  • Thanks. I just thought people here might meet the same problem. I don't understand why the VAR model in R allows no zero value in variables. – user2502240 Jul 15 '14 at 18:33

0 Answers0