The scope of scipy / numpy is such that I think you'd be hard pressed to to prove that 'scipy is numerically stable' or something to that extent - instead you'll need to look at each algorithm (or class of algorithms) independently. Much of the scipy / numpy functionality comes from a relatively thin wrapping of underlying c or fortran libraries. These libraries e.g. lapack/blas (or Atlas or MKL) are generally industry standard libraries with well documented numerical stabilities. I think the best strategy would be to try and find out which of the c or fortran level libraries your code uses and go from there. For routines which aren't a simple wrapping of a library call, there are often references to papers describing the algorithms in the documentation or comments. The unit testing code might also be a good place to look. cheers, David ________________________________ From: Marcel Caraciolo <caraciol@gmail.com> To: scipy-user@scipy.org Sent: Wednesday, 18 January 2012 8:58 AM Subject: [SciPy-User] Information about the Numerical Stability of Scipy/Numpy Hi all, I am studying scipy and numpy and  I decide to write some routines with those packages for a paper submission in a scientific congress. The problem is the validation of the results and experiments I have to show that the libraries that I used (in this case Scipy and Numpy) provides numerical stability, otherwise the chances that my article be approved will be decreased. Any further information in docs or the website about this topic ? Regards, -- Marcel Pinheiro Caraciolo M.S.C. Candidate at CIN/UFPE    http://www.mobideia.com    http://aimotion.blogspot.com/ _______________________________________________ SciPy-User mailing list SciPy-User@scipy.org http://mail.scipy.org/mailman/listinfo/scipy-user