[6278] in s-news-athena
Excel
daemon@ATHENA.MIT.EDU (Frank Harrell)
Fri Jan 27 10:10:25 1995
From: Frank Harrell <feh@biostat.mc.duke.edu>
Date: Fri, 27 Jan 1995 09:27:01 -0500
To: s-news@utstat.toronto.edu
It's strange to see S-Plus mentioned in the same paragraph with Excel.
One of the most basic statistical operations is becoming the generation
of a scatterplot with a loess-smoothed trend line added. I'd like
to see someone do this with Excel.
There's a basic problem among many analysts - that something should be
easy to use to be worthwhile. It makes me recall the saying "if you
design something so that an idiot can use it, only idiots will".
This same argument applies to the question of whether it's worth the
time to learn the Emacs editor. (The answer is YES!).
Another example is the emergence of menu-driven stat packages in the
last few years. See how far they get you when you want to not
assume linearity for a predictor variable in a logistic regression.
Does the package allow you to generate nonlinear B-spline or
truncated power basis functions to add to the model?
