I have a question on VECM model. I have a set of variables I had planned to include in my VECM model where one particular variable may be trend stationary (@ 10% s.l. by ADF test) while the rest are definitely not. I know the first step is to find the lag length of the VAR mode, But I found that if I fit the VAR model with lag chosen by AIC on all the variables, I get autocorrelation in the residuals. However,if I remove this variable that may be trend stationary, I do not find autocorrelation in the residuals. Examining the residuals specific to this variable shows autocorrelation in several lag terms, whereas the residuals for other variables have no autocorrelation at all. Does this mean I should remove this variable from the VAR model?