MATLAB: Does NARX use the current value of the eXogenous input

Question

I understand that to predict y(t) using NARX, you will use the previous values, delays, of the target time series, say y(t-1) and y(t-2), in addition to the previous values of an external time series, say x(t-1) and x(t-2). Does NARX uses x(t) to predict y(t)? and how to prevent it from doing so?

Answer 1

The default input delay ID = 1:d, (d=2) does not use the current value.

However, if you wish to use the current value, just add 0 to ID: ID = 0:d

On the other hand, adding 0 to FD is only allowed in the OL configuration. Therefore, I see no reason to use it.