s-news
[Top] [All Lists]

[no subject]

To: s-news@wubios.wustl.edu
Subject:
From: Wim Kimmerer <kimmerer@sfsu.edu>
Date: Fri, 11 Nov 2005 13:10:27 -0800
Cc: "Re:"@wubios.wustl.edu
I don't know the reasoning behind not including Bartlett's test, but in general the philosophy behind Splus has followed that of Cleveland and other authors who advocate graphical, exploratory approaches to statistics. If you look at a normal probability plot of your residuals, that should tell you all you need to know about the degree of normal-ness of your data. Besides, it makes a huge difference what you are going to do next - if it is a balanced ANOVA with no post-hoc testing you can get away with more deviation from normality than Bartlett's test would likely let you get away with. So a careful look at diagnostic graphs (which are easier to do in Splus than anything else I know of) by an informed analyst beats rote application of procedures hands down.

Regarding GUI vs. no GUI, it can be easier the first time or two, but the command line is much quicker once you learn how to do something and don't have to go pointing and clicking all over the place.

Wim



"Overstreet, Jason (FL51)" <Jason.Overstreet@honeywell.com> writes
...
know that when I do a test of equal variance, I use Minitab.  When I wrote
Insightful, they suggested that I write Bartlett's test from scratch.  I
wrote it from scratch just to see how hard it would be which took a few
hours out of my life, but this is an example of such a basic test, that
Insightful should have included on the GUI.  If I were the CEO of
Insightful, there would be no way that I would let Minitab have a built in
canned capability that Splus doesn't have.  Convenience is the only thing
that Splus has over R and if they want to stay in business, then they better
not forget this.
...


<Prev in Thread] Current Thread [Next in Thread>
  • [no subject], Wim Kimmerer <=