Re: Statistica

Paul F. Velleman (pfv2@cornell.edu)
Mon, 22 Aug 94 17:42:03 EDT


In article <199408150004.KAA00423@eos.xx.rmit.EDU.AU>,
reece@eos.xx.rmit.EDU.AU (John Reece) wrote:

> 7. There were problems with early versions of Statistica, and there was
> lawsuit filed against the makers of Statistica by Systat. I'm told
> (admittedly by the Statistica people) that all bugs have been ironed out
> and that the lawsuit was thrown out of court.

To the best of my knowledge (including recent conversations with Lee
Wilkinson, the President of Systat) there was never a lawsuit filed against
Statsoft, the makers of Statistica by Systat. Wilkinson published a
pamphlet that presented very convincing evidence that the designers of
Statistica had plagiarized extensively from Systat (including such internal
evidence as their misunderstanding an example in the Systat manual,
thinking it was the definition of a new method. The example appears only in
that one place, so the Statistica "method" must have been copied from that
source.)

Wilkinson argued further that there was much internal evidence that the
designers of Statistica are not statistically sophisticated. He noted a
number of bugs -- not to claim that statistica was buggy, but rather to
point out that no responsible, knowledgeable firm would ship such a
product. (e.g. residuals that were not uncorrelated with factors, and other
gross errors.) Statistica's response was to smear Systat with inuendo and
half-truths, and by finding a few places where Systat's calculations broke
down.

It is true, as far as I know, that the bugs in Statistica pointed out by
Wilkinson were fixed (as were the bugs in Systat.) The point that was
missed by many readers was that the Statistica bugs were fundamental and
indicated that Statsoft was either incompetent or irresponsible, while the
bugs in Systat were the sort that occur at the edges of comptuational
precision where all calculations break down and sometimes the test to catch
the situation fails to kick in in time.

My take, from the outside of the debate, is more in terms of the style of
both firms. StatSoft regularly attacks its competitors with smears, lies,
and half-truths. (If they told John that the suit was "thrown out of court"
that would be par for the course. No suit ever was filed.) I find this
behavior reprehensible and question whether professionals behave this way.
In a market that includes many fine products, I think that it is better to
work with responsible professionals.

-- Paul Velleman