I am estimating a logit model. Several of the predictors are time series. Rather than differencing the series for stationarity I would like to test whether the series in predictors are cointegrated. Is it appropriate to do that? If so, what STATA operation is best used?