I did an ADF test on each stocks of the S&P500 to see which one were stationnary. I did the test using a sample of 30 days by 1 minute interval.
By curiosity I expanded the test to 60 days, and the results changed, I expanded again to 90 days and the results changed again.
For some stocks I took a period where I knew the ADF test would show stationnarity and then added or substracted just one period, this was enough to change the results from stationnary to non-stationnary.
I don't understand why even one period changed everything.