URADF with GETS-procedure

[Anna]

Is it correct that the general to specific procedure in uradf does not correct for the different sample size in each step? Or did I misread the code? (Reading is not my strength anyway). As far as I know, Ng and Perron have argued that such a correction is important, haven't they?

[TomDoan]

For the IC's, it's important to use the same range for all the models, since you're comparing all lags with each other. Ng and Perron's warning was due to them having seen that programmed wrong in some cases.

With GTOS, you can go either way - a common range, or an increasing range as you drop lags. That's because you don't compare 1 lag with 8, only 8 with 7, 7 with 6, etc.