Dear members,

I was performing the DF test on a AR(1) model,

I was firstly trying dfuller nilemax, regress lags(1), however this gives me an augmented DF statistic.
Therefore, I tried dfuller nilemax, regress lags(0) and this time I get the DF statistic.

I am a bit confused here: should AR(1) include one lag difference? If yes, then why I need to choose 0 lag difference to get the DF statistic?

Many thanks