In order to de-trend a trend stationary variable, I conducted a linear function (variable = intercept + trend) and use the residuals as “detrend” series, is that correct?
Does the dependent variable need to be stationary for an ARMA model? If it’s trend stationary, the dependent variable should be the variable itself, detrended series (residuals), or the 1st difference?
Which correlogram should I look at when I estimate the model? The correlogram of the variable or of the residuals…?
You guys see ARMA models in L1 now? Wow, plenty of changes.
If your series is not covariance stationary, then use 1st difference and watch again. Don’t forget that the interpretation of coefficients has also changed.
If you want to use OLS method (or similar) for regression calculation, then all the variables in the model (independent and dependent) must be covariance stationary, so yeah, also for ARMA models.
As far as I know, the “correlogram” is only for residuals.
