Evaluation of autocorrelation

[Bablowski]

I use ARDL model for my estimation, dependent and independent variables all max. 2 lags. Now I want to check the residuals and perform its diagnostics. When I use Correlogram Q-Statistics for checking autocorrelation, EViews asks me to choose the number of lags to include. What number of lags should I include? 2, when I estimate variables with 2 lags or more lags? How many and why?

Thank you for any help.