Unit root tests

[snsburnttoast]

Hi all,

I am using unit root tests (namely PP and KPSS) for monthly time series analysis. I have a whole sample of 1998m01 to 2015m12 and also am using a subsample of this from 1998m01 to 2008m08. I am going to then go on to do Johansen Cointegration and VAR estimation after so I need to determine what order my series are integrated in. For further information they are the following variables: stock market index, cpi, interest rate, exchange rates, industrial production index, money supply (2), and oil prices. Note these are all in log form. I'm doing an analysis of these on four countries.

My question is in regard to determining unit roots for the analysis. Do I only have to do unit root tests at the whole sample level to judge whether they are I(0/1/2) for analysis, or do I do them at both the subsample and whole level for both parts of my post unit root analysis (subsample/whole sample)? This is as my results from the subsample and whole sample differ for the countries.

Thank you in advance for any help.
James