Time series


Fortran must die
For the augmented dickey fuller SAS generates a PR<Rho and PR<TAU. These are test of significance, I don't understand which you should use to test the null (that is what are the substantive difference that would lead you to chose one of these to test the null over the other).

A second question, which has nothing to do with SAS but is related, is why you chose one of the 3 augmented dickey fuller possibilities. There are the zero mean, the constant and the linear trend. What in you assumptions about Stationarity leads you to test one of these over the other.


Fortran must die
It is in several, but here I use PROC AUTOREG.

This code generates it [I don't have a handy data set so this comes from the Milhoj book on time series].

PROC AUTOREG DATA =sasmts.quarterly_milk PLOTS (UNPACK) =ALL;
MODEL COWS = /Stationarity =(ADF);