Date: Fri, 8 Dec 2006 05:55:14 -0800
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
Subject: Re: OT: Understand world geography -- in color
Content-Type: text/plain; charset="us-ascii"
Jerry Davis wrote:
> toby dunn wrote:
> > As my old prof use to say 'You cant prove squat with a graph, its the
> > numbers, math, and stats that matter'. When you make a graph the graphics
> > should be the bridge between the numbers and something people quickly
> > recognize. When I look at a graph that has one picture that is big and
> > another small, supposing I recognize the picture, I can make some very
> > limited conclusions,. But I would also want to know how big are they, what
> > is the differents in hard numbers and percentages of teh whole. Once you
> > drop the numbers and distort the thing the person was suppose to recognize
> > the graph becomes crap.
> If someone wants to distort the facts, it is just as easy to distort a
> numeric based presentation as graphical based one. If you dislike
> graphics, fine. That doesn't mean that they are inherently misleading,
> cartograms included.
> Jerry Davis
> Experimental Statistics
> UGA, CAES, Griffin Campus
This might be an opportune moment to mention a couple of things:
1) The famous quote, attributed to Benjamin Disraeli: "There are three
kinds of lies: Lies, damned lies, and statistics."
2) A book that I always have handy--"How to Lie With Statistics", by
Darrell Huff. IMHO, this book should be mandatory for anyone who
creates or uses statistical analyses. It was originally written in the
early 1950's, but is probably even more valid today than then.
As noted, any kind of statistical analysis can be distorted (willfully
or otherwise), whether tabular, graphical, or whatever. Both the
analyst and the interpreter have to know exactly what the objective is.