QUESTION:
The attraction of
Transfer Function modeling is that, apart from getting a superb fit and being
confident
of forecasts, we gain a causal insight into
the relationship between the input and output. The mechanics of
some of your internal adjustments are also of
interest and I presume these are your own invention rather
than following something in the literature
(which is apparently not yet complete). Is all this a trade
secret??
ANSWER:
To the contrary,
AUTOBOX's procedures, approaches and implementations are all in the public
domain,
well nearly all ! The identification of the
relationship between time series and an iterative approach to the
"best model" were all outlined in
the 1968 text by Box and Jenkins. Various dissertations and state-of
-the art research have
winnowed down the number of things that can go wrong. AUTOBOX is very
aggressive in incorporating these ideas.