Ken wrote:
> I am a big MINITAB fan, but I'll be as unbiassed as I can be. . .
>
> When I was looking for a general statistics package to provide for engineering
> use (basic stats, regression, GLM, reliability/survival analysis, DOE, etc) the
> only packages that had the tools I needed were Statistica, JMP, MINITAB, and
> StatGraphics. I spent a lot of time playing with each of these.
>
NCSS has had all these capabilities for the last 10 years. You must have been
looking a long time ago.
>
> I VERY seriously considered Statistica, it has a very good graphics engine, but
> I found it much too hard to use. I found it could do everything, but I was
> always struggling to get it done. It made my brain hurt. I seriously encourage
> you to download their demo - although it is probably so hobbled it may be hard
> to use.
>
> StatGraphics has a very old windows environment, so I ruled it out pretty early.
>
> JMP 4 is a pretty good stat package, but I've always found JMP to have an odd
> environment, especially for someone who knows what they want to do (what tests
> to run, etc). Their Fit Model tool drives me nuts. JMP 4 is slightly improved,
> esier to find stuff, but not much better. It does have a slightly better
> survival tool, but it is limited - right censor only - no interval censoring or
> plots that I can find. Do download the JMP 4 demo - it is hobbled, so you can
> only use their data, can't print, etc - at www.jmpdiscovery.com. It makes it
> hard to evaluate fully.
>
> As I said, I really like MINITAB, and strongly recommend it. It has very
> powerful GLM, , time series, regression, Survival, and basic stats tools. It
> also has good graphics, although not as nice as Statistica. They have a fully
> functioning demo (the only "hobble" is it stops operating after 30 days - and
> they can even extend that if you call them) at www.minitab.com.
>
> If your objective is to do statistics with less sweat, then MINITAB should be
> looked at.
> If your objective is to get slick graphics, but work hard, then check out
> Statistica.
>
If your objective is both low-sweat statistics and good, fully-customizable
graphics, check out NCSS.
>
> I think both JMP and MINITAB are about the same price - I think Statistica will
> be quite a bit more expensive.
>
I'm a big fan of NCSS, although I had to use Minitab and SPSS when teaching an
established curriculum which used the school's mini-computers. I own personal
copies of all three as well as S-Plus and earlier versions of Statistica and
Statgraphics Plus. I have not found many problems that NCSS could not handle with
usually greater speed and much more versatile graphics. At a price of around
$300(US), it is a super-bargain and, I believe, a superior product. (Anecdotal
story: A few years ago the NCSS probability calculator disagreed with SAS's
hypergeometric probability routine output. We contacted SAS and they found they had
a programming error - since corrected.)
Often, personal preference boils down to that which one is most used to using.
However, I have found that the NCSS help files, which include an integrated tutorial
for every procedure, are extraordinarily clear and helpful (as help files should be
but often aren't). Additionally, there are no add-on modules, the basic package is
complete (although it does interface directly with PASS - Power Analysis & Sample
Size - program.) You can download 30-day fully functional trial at www.ncss.com
(where else?).
Tom Gatliffe
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
