Hi I have some issues with the stationarity condition for some variables in a time series analysis with VARMAX. As you can see the autocorrelation and partial autocorrelation seems to be negative in most of them when the variables are logdifferenced (dl in front of variable name). When searching for this pattern in ACF and PACF this could be because of overdifferencing. Only taking log (l in front of variable) although looks much worse taken stationarity in consideration. What is the right way to ensure for stationarity for these types of variables?

Similar questions and discussions