Aff Bias Update: ans Dove
Wed Sep 10 07:51:27 CDT 1997
I think that in the post below the problems of trying to do analytical
objective research in the human realm has raised its ugly little head.
shuman argues against looking at a wider set of data because it will be
harder to control for all of the variables and that the added data will
contaminate the pure cause effect relationship that he is lloking for.
So he willsubjectively select his data leaving out important alternate
causes to the probleme that may exist. from these subjective decisions
he will arive at "objective" data that will tell us all the "Truth"
about the aff/neg split. With this "objectively" arrived at Truth we
will no longer need to rely on the subjective stories of that scarry old
BEar and his 20 years experience. Yeah right.
From: Terrance Shuman
To: EDEBATE at LIST.UVM.EDU
Sent: 9/9/97 5:13:25 PM
Subject: Aff Bias Update: ans Dove
In <Pine.SUN.3.95L.970908220229.21056D-100000 at ahnnyong.cc.columbia.edu>,
on 09/09/97 at 10:59 AM, Alan Dove <ad52 at columbia.edu> said:
>I'm thinking through the target and wondering what might cause such
>a phenomenon if it is shown to exist. Nobody should read my posting
>and say "Oh, there's the reason for the bias."
I understand. However, one of the reasons I like looking at
high-skill level samples is that it cuts down on the need to worry
about controlling for things like "how hard the Negative is to
learn," etc. If, indeed, Neg is a more difficult side to master,
that complicates the examining of data from teams at the middle
or the bottom of the skills ladder, since we can't be sure of how
much the Aff/Neg splits are attributable to the bias we are trying
to locate, and how much to the teams' collective lack of skill at
upholding the Neg side. (It is an assumption that we can eliminate
this by "matching" skill levels even at those lower thresholds. I
don't think just matching the skill levels solves this problem,
>I'm just throwing in some theories to point out that there
>are reasons to do more thorough studies of A/N win ratios. In
>particular, I would like to know as closely as possible what the
>numbers are at different levels, and whether or not topic choice
>makes a difference.
It would certainly be interesting to look at year-by-year results
and correlate them with resolution "types," but I don't know how
you could do this scientifically. There doesn't seem to be much
agreement about which resolutions "favor" Aff and which ones
"favor" Neg. Those discussions *are* colorful, though.... ;-)
>The question is now what kind of data we need. I think an analysis
>of prelim rounds at the top and bottom of the bracket would be
>constructive, because biases which might not show up in elim rounds
>could show up there. My purpose in proposing mechanisms was to
>illustrate ways in which a bias could appear at the lower end which
>is absent from the upper end. Until we look, we won't know.
This is certainly true. But the question remains, what do we
"know" if the bias shows up at lower levels but not at upper levels?
If results are skewed toward Aff at the mid- to lower levels of
skill, how does this help us decide if the famous bias exists?
Examining the results of the top performers seems the best way to
make sure we bag the critter we're after, one way or the other,
with no wiggle room.... ;-)
>Always try to think of the next experiment.
Man, my brain is still cramping from thinking about *this* one!
Everything I know about this stuff I learned from reading Bill James's
baseball books and John Allen Paulos's A MATHEMATICIAN READS THE
Bishop LeBlond Memorial High School
St. Joseph, Missouri
More information about the Mailman