I am happy you enjoyed the blog post. Cointegration can only occur between non-stationary series. Therefore, it is invalid to consider cointegration if all or any of the time series are stationary.

Thanks for the amazing post on cointegration.

I have a question on what needs to be done after performing a Johansen's test for cointegration. So lets say I have 4 time series on which I am doing the Johansen test, and the results indicate that we reject the null hypothesis at r <= 1, so is it correct that we have to make a linear combination of 2 variables out of the 4 to make a stationary series? We have the 4x4 eigenvector and we also have the highest eigenvalue (which corresponds to the first column of the eigenvector, right?). So, is it the first 2 of the 4 timeseries that we chose to make our linear combination with the weights as explained in the quantstart article (https://www.quantstart.com/articles/Johansen-Test-for-Cointegrating-Time-Series-Analysis-in-R/)? Seems counterintuitive, as the 4 time series could have been entered in any order, right? In that case, how do we know which 2 variables to choose to make the stationary linear combination?

Thanks
Vin