Rich Ulrich wrote:
>I am not positive, but
>I think I would have objected to "equal % change"
>as =proportionate= by the time I finished algebra in high school.
>
>I know I have objected to similar confusion, on principled
>grounds, since I learned about Odds Ratios.
>
>I suspect that the original sample was small enough that
>the apparent difference in ORs was not impressive.
>--
>
I too think that the odds ratio is the appropriate way to present the data, but
after looking at these results, I can appreciate why the Gallup organization
didn't do so.
The data on racial favorability ratings which Gallup called 'proportionate' not
'disproportionate':
GWBush favorability pre- and post-9/11
Pre-disaster Post-disaster OddsRatio
White (odds) 60% (1.5) 90% (9) 6
Black (odds) 33% (0.49) 68% (2.1) 4.3
GHBush favorability pre- and post-Gulf War
Pre-disaster Post-disaster OddsRatio
White (odds) 64% (1.8) 90% (9) 5.1
Black (odds) 33% (0.49) 70% (2.3) 4.7
Unless I'm missing something, a logistic regression analysis with disaster and
race coded as dummy variables with an interaction term (disaster x race) would
have a sign for the interaction coefficient indicating that the odds of white
favorability for Bush after a disaster increased more than the increase in
black favorability for Bush after a disaster. This is NOT what I'd expected
after looking at the raw percentages, which to me indicated a
disproproportionate increase in black favorability for the Bushes after
disasters.
Gene Gallagher
UMASS/Boston
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at
http://jse.stat.ncsu.edu/
=================================================================
