EViews.com

Hi guys this is my first post on the forum. So im currently studying time series analysis and have a massive assignment due on it. the problem is that im not too sure if i have understood the order of methodology correctly and was wondering if you guys could help out.
So lets say you have some series Y(t) which you want to forecast.
Now is the first thing to do is to determine the model structure (i.e. lag lengths) i.e. AR(p), MA(q) or ARMA(p,q), and then estimate the model (after adjusting for trends etc) after this has been determined do you then carry out unit root tests(i.e DF or ADF) on the AR component of the selected model(even if the model is ARMA) and then forecast if the series turns out to be stationary

THanks for the help .

Unit roots tests come first, then, when the j-order of integration of the variables is determined (in case of variables that are stationary in differences), then you fit an ARIMA(p,j,q).