Re: Study about accuracy of statistical software, incl. Octave

[Francesco Potorti`]

>
>On 22-Mar-2009, Ivan Sutoris wrote:
>
>| Through R-help mailing list, I've stumbled upon an upcoming study
>| comparing accuracy of free statistical software (R, Python, Ox and
>| Octave), which I thought may be of some interest here. For Octave
>| (authors used version 2.9.12), they report some problems with accuracy
>| when computing correlation coefficient (corrcoef), quantiles for some
>| distributions (tinv, chi2inv) and for linear regression (regress from
>| Octave-Forge). Here is the reference and link:
>| 
>| Almiron, M. and Almeida, E. S. and Miranda, M.: The Reliability of
>| Statistical Functions in Four Software Packages Freely used in
>| Numerical Computation. (to appear in) Brazilian Journal of Probability
>| and Statistics, Special Issue on Statistical Image and Signal
>| Processing
>| http://www.imstat.org/bjps/papers/Almiron.pdf
>
>Writing a paper complaining about problems in free software is a
>ridiculous thing to do.  Instead of writing a paper, they should have
>reported the problems so that they might be fixed.

I wouldn't say so.  Writing a paper like that has the benefit of letting
people know about what's available and how tools compare, which is
important when deciding what to use.

It also benefits the developers, because it can make them focus on
problematic aspects of the software, more or less like a bug report.
Sure, it would have been nice if they sent an announcement of the paper
on the bug list, but it arrived here nonetheless.

-- 
Francesco Potortì (ricercatore)        Voice: +39 050 315 3058 (op.2111)
ISTI - Area della ricerca CNR          Fax:   +39 050 315 2040
via G. Moruzzi 1, I-56124 Pisa         Email: address@hidden
(entrance 20, 1st floor, room C71)     Web:   http://fly.isti.cnr.it/