The reality is that for every different FFT algorithm (and probably every
package you don't write), you are required to determine what happens to your
input and what the output is. FFTW is relatively straight-forward in this
respect. At least the parts I've used.

In article <1168018842.035090.317130@v33g2000cwv.googlegroups.com>,
stevenj@alum.mit.edu wrote:

Dear all,

I am sending you another list of articles accepted in this year.
All accepted articles can be found at

http://www.stat.sinica.edu.tw/statistica/

(Please click on Pre-prints.) The full-text may not be available at the
above site for those still under technical editing or final revision
(by
the authors). Please contact the corresponding authors for a first
hand reading of their work. Please also keep me informed of your
questions on our preprint articles (karen at stat.sinica.edu.tw) .

With best wishes for the New Year,
Karen Li, on behalf of the Co-editors
Editorial Assistant

-------------------------------------------------------------------------------------------------------

SS-06-031
Confidence intervals for the long memory parameter based on wavelets
and
resampling
Pier Luigi Conti

Hello,
I want to solve Poisson's equation on a 2d, n*n, rectangle with
periodic b.c. How should I set up the matrix of coefficiets in finite
differene? Ku=b, K=[-2 1 0 0 ...0 0 1; 1 -2 1 000.....000 ; ......; 1
000....0000 1 -2 ] is singular, have 'n' zero eigen values. (Note: I
am only allowed to use finite difference, and boundary conditions are
PERIODIC).

Thank you,
Farshid