Thread Subject: What's the problem of the following function I defined?

function y=test(N)
Sum_tp=0;
for i=0:N-1,
     Sum_tp=Sum_tp+nchoosek(N-1,i).*((-1)^i)./(i+1)*exp
(i+1)*expint(i+1);
end
y=Sum_tp;

The function y=test(N) I defined above shows strange
results for inputs larger than about 50.

I guess there happed some resolution problem, probably in
nchoosek(n,k) function, but, I'm not sure.

I would appreciate it if you gave me clear explanation.

"HG Cho" <ddol1129@gmail.com> wrote in message
<fm7ioe$qie$1@fred.mathworks.com>...
> function y=test(N)
> Sum_tp=0;
> for i=0:N-1,
> Sum_tp=Sum_tp+nchoosek(N-1,i).*((-1)^i)./(i+1)*exp
> (i+1)*expint(i+1);
> end
> y=Sum_tp;
>
> The function y=test(N) I defined above shows strange
> results for inputs larger than about 50.
>
> I guess there happed some resolution problem, probably in
> nchoosek(n,k) function, but, I'm not sure.
>
> I would appreciate it if you gave me clear explanation.
>
>

Please refer to nchoosek function

>> help nchoosek

NCHOOSEK Binomial coefficient or all combinations.

    NCHOOSEK(N,K) where N and K are non-negative integers
returns N!/K!(N-K)!. ....
   When a coefficient is large, a warning will be produced
indicating possible inexact results. In such cases, the result
is only accurate to 15 digits for double-precision inputs, or
8 digits for single-precision inputs.

and factorial function
>>help factorial

 FACTORIAL Factorial function.
   ....


Since double precision numbers only have about 15 digits, the
answer is only accurate for N <= 21. For larger N, the answer
will have the correct order of magnitude, and is accurate for
the first 15 digits.

Anh Huy Phan
RIKEN - BSI