This message is posted at the request of
StatSoft, Inc.. StatSoft does not have
direct access to the Internet.
*******************************************
RESPONSE TO
MR. SVEN BLANKENBERGER
We found the recent posting by Mr. Sven Blankenberger
(concerning the comparative review of statistical
packages in c't Magazine) very surprising and need to
clarify a number of inaccuracies and important
omissions in his note.
We appreciate the fact that Mr. Blankenberger has
identified himself as a person who is associated with
Systat. However, we believe that the content of his
note goes far beyond a "bias" which could be explained
just by "differences in interpretation" of some
ambiguous facts.
Mr. Blankenberger states that c't Magazine is in his
opinion the best German computer magazine. We agree,
c't is an excellent magazine; however, this appears to
be perhaps the only statement in his letter that does
not include or imply an error or misinterpretation.
1. The statement that the article was not a review
but "a report just describing some features" is highly
inaccurate. The article is one of the most
comprehensive professional publications evaluating
statistics software ever published in any German
language magazine! It was clearly designed to provide
a comparative evaluation of the products, it is very
long (10 pages), very thorough, and as we know first
hand, highly regarded by the German market place.
2. The statement that all packages were reviewed by
one reviewer is accurate, however, it represents
rather a strength than a weakness of the review:
Because of this fact, precisely the same evaluative
criteria were applied to all products and thus the
ultimate level of consistency of evaluations could be
achieved.
3. Mr. Blankenberger's statement that STATISTICA did
not receive the highest overall rating is WRONG.
There were in fact four dimensions. Two of them:
"General Functionality" (Funktionsumfang) and the
"Quality of the User Interface" (Bedienbarkeit) were
the most important not only (1) because they
practically exhaust what is crucial in the evaluation
of ANY software product, but also (2) because the
other two (indisputably less important) dimensions
("Manual" [Dokumentation] and "Expandability via Add-
on Modules" [Ausbaufahigkeit]) were simply NOT
APPLICABLE to some of the reviewed software packages.
Specifically, some software, like STATISTICA, offers
no add-on modules simply because ALL potential add-on
modules are included in the product at no extra cost.
To illustrate this point, compare how much a user has
to spend for SPSS or Systat add-on modules in order to
be able to perform Reliability/Item Analysis,
Canonical Correlation, Survival Analysis, Quality
Control Charts, Experimental Design and other types of
analyses which are available in STATISTICA at no extra
charge. StatSoft does not sell software piecemeal.
Even the SEPath module, a new, comprehensive
implementation of Structural Equation Modeling
techniques (equivalent to LISREL sold by SPSS
separately for $495) will be included in STATISTICA at
no extra cost. In this context, Mr. Blankenberger's
arithmetic of subtracting points for not charging
extra for these modules is highly unfair and
misleading! The other dimension that was not
applicable to all packages was "Manual" which was not
available at the time of the review and thus could not
be evaluated. This fact was stated explicitly in the
footnote in the article (on page 185; did Mr.
Blankenberger miss this footnote?).
4. STATISTICA was the only package which received the
highest rating (that is, two "pluses") on the two,
indisputably most important dimensions ("General
Functionality" [Funktionsumfang] and the "Quality of
the User Interface" [Bedienbarkeit]), and--let me
repeat--the only ones which were applicable to all
evaluated software. Now, even using Mr.
Blankenberger's method of translating pluses and
minuses into signed integers, the correct comparison
of the four major contenders is as follows:
+4 STATISTICA
+2 SPSS
+3 Systat
+2 SAS
5. It is not true that there were "serious errors"
that were corrected in the "following issue" of the
magazine. c't Magazine did not admit to any "errors;"
there were only two minor clarifications (concerning
SYSTAT's underflow option and internal precision, and
the upcoming Beta (pre-release) version of UNISTAT
2.0), with no bearing on the overall outcome of the
review.
6. We strongly suggest that Mr. Blankenberger ask a
sample of readers of the review article in c't
Magazine for their understanding of the results of the
comparison. Note that StatSoft of Europe is eagerly
distributing this review, and has been very pleased
with its impact. Numerous customers placed their
orders based solely on their clear understanding that
STATISTICA won, and one can hardly imagine how this
review could be read in any other way. Does Mr.
Blankenberger really believe that one should subtract
from the maximum number of points (+4) received by
STATISTICA (and ONLY by STATISTICA) on the two
objectively most important dimensions ("General
Functionality" and the "Quality of the User
Interface") any points because its competitors charge
extra (i.e., offer add-on modules) for the same
functionality? Does it really sound like an unbiased
way of reading the comparative table?
7. Finally, there is no word "successfulst" in the
English language; the correct translation of the
German "gelungenste" is "the most successful" or "the
best" -- therefore the translation which appeared in
the note by Dr. Czyz "the best Windows implementation"
is precise and accurate.
We invite all interested parties to verify the facts
presented here in the text of the review: c't
Magazine, 3/94.
Bernd-Uwe Loll
Managing Director
StatSoft of Europe
Hamburg, Germany