Author: Roy Eassa
Date: 16:35:22 06/06/02
Go up one level in this thread
Let's see if I can paraphrase a concept I think you've been making:
If one takes an EXISTING set of results, one can find statistics that support
almost any theory. ("There are lies, damn lies, and statistics.") There are
essentially an unlimited number of statistics that one could choose to focus on.
In order for a statistical experiment to be scientifically valid, all variables
need to be clearly controlled and/or spelled out BEFORE the experiment is run.
Also, exactly which statistics will be collected, and how, must be decided
before the experiment is run (in order to avoid selection of those statistics
that are attractive after the fact).
Does that make sense? Is that approximately what you've been saying?
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.