1

I have 2 questions on the Westerlund cointegration test. I have panel data from year 1995 to 2015.

1) How do you decide the lag? So for example, if i were to include 3 lags, it says i need at least 22 observations But i have some series that have missing data, so it gives me an error msg saying the series do not contain sufficient observations.

Same when i pick 2 lags, or 2 leads, there are some series whose observations are not complete. Is picking the right lag/leads very crucial in testing for cointegration?

2) i get the following results when i run the test. All tests are statistically significant except the Pt statistic. How do i go about this??

enter image description here

kjetil b halvorsen
  • 77,844
Olivia
  • 119
  • I would change in Options model selection criterion to Hannan-Quinn and then Model Selection Summary go for Criteria Graph. The shorter is the bar for the particular model, the better is this model. – Pipi Jul 25 '21 at 18:53

0 Answers0