These steps should help you pass through the teething stages on using ARDL, depending on what software you're using (I use Stata):
1. indicate that you want to perform a TS analysis, for Stata, you have to type "tsset year";
(2). Choose optimal lag length for the model,
(3).Undertake unit root testing - ADF, DF-GLS, Correlogram, P-P etc.,
(4). Perform ARDL level regression,
(5). Next is the ARDL Error correction representation,
(6). Bounds test for cointegration,
(7). Post-estimation tests - DW, BP, BG, Ramsey, JB, CUSUMSQ, Unit circle etc
Note: In most cases you have to perform the cointegration test before running the ARDL regression, but the reverse is the case in Stata where the command for Bounds testing for cointegration is generated after the ARDL error correction regression.
Also, for unit root testing, it does not matter whether your variables are I(0) or I(1) series in ARDL unlike what obtains under the VAR model. But we still test for unit root to ascertain that none of the variables are integrated of order 2, that is, I(2) or higher.