To apply the T transform, the assumption that all
had to be made. Here we present a simple theorem that expands the region of validity of the various expressions derived in this paper to the region where any of the
may be non-negative. We present the theorem for the single subset sum case only, although the multiple non-overlapping subset case and the contained overlap case may be handled in an almost identical manner.
: If
,
and
,
then
![]()
Proof: Note that
implies that there is an integer q>0 and an
such that
. Thus
may be rewritten as
![]()
where
, the Kronecker delta function. Iterate this operation q times (removing one power from
and summing with an increased count vector each time) to find
![]()
Simplify this to yield

where the vector
has nonnegative integer components summing to q with
for
. Since
, evaluate the integral
using theorem 12 with k=1 (noting that
and
increase by q due to
being added to
) to find

Now, we put
into closed form by noting that it is the discrete convolution product of the functions of
of
given by

Apply the Z transform convolution theorem (see appendix 9.6.2) to find

Note that
![]()
for
and substitute for the Z transforms to find

Substituting this result in (*) and simplifying leads to the desired result. QED.
We resort to analytic continuation in the non-contained overlap case.