How to find the order for ARMA model? I have a problem finding the order with the ACF and PACF plot, below is it. 
First I think they can be considered as tails off gradually because they are abnormal, then I set AR(1) from PACF and MA(1) from ACF, but if I use auto-select order then it gives me higher order. I know the ACF&PACF are just guessed, but I still want to know how can we deal with the plots in this form? And how to set the possible order properly.
Thanks!
 A: When considering AR or MA or ARMA models, you already mentioned correctly we look at everything outside the error bans. Everything close to zero, "within the errors bans" will be not significant as a lag order.
If we take for example the PACF, for an AR Model, such a structure, as in your image, would indicate that,
lag 1, 2, 4, 6 and 15 might be good candidates to fill the model:
See also here at 7:25
https://www.youtube.com/watch?v=5-2C4eO4cPQ.
The same method goes for MA and ACF, in ARMA, however, it is mostly the case, that the highest order is chosen, and this is probably what your auto select did:
5:40
https://www.youtube.com/watch?v=HhvTlaN06AM
And the author states, what is obviously correct:

..., the order of the AR or MA part is based on the last
significant lag in the PACF / ACF respectively.

by ritvikmath
So in your case the ARMA model would be (16,16) order by the look of it, although they are slightly over the error bans. But this "last signifcant" rule explains your auto select
