"Neil W. Henry" <[EMAIL PROTECTED]> wrote:
> Date: Thu, 16 Nov 2000 14:49:09 -0500
> From: "Neil W. Henry" <[EMAIL PROTECTED]>
> Subject: Re: NY Times on "statisticians' view" of election
>
> Paul Thompson wrote, speaking of "caustic jerks":
>
> > Herman Rubin wrote:
> > >
> > > You may be making a Type 3 error. Remember, the null
> > > hypothesis is always false.
> > >
> > > Those who voted for Bush are more likely to be literate,
> >
> > This is the kind of offensive, stupid comment that belongs on political boards.
> >
> > Anyone who makes such comments without documentation, without evidence
> > is simply not a scientist. He simply is not even an educated person.
>
> Rubin's is not a very controversial statement. I would think that most readers
> of this newsgroup not only agree with it, but have access to documentation of it.
> Here's a table from the 1996 General Social Survey of American adults that shows
> that partisan Democrats score lower on a short (10 word) vocabulary test, on the
> average, than partisan Republicans.
>
> NUMBER WORDS CORRECT IN VOCABULARY TEST
> POLITICAL PARTY AFFILIATION Mean N Std Dev Grouped Median Std. Error of Mean
> STRONG DEMOCRAT 5.83 263 2.22 5.81 .14
> NOT STR DEMOCRAT 6.02 365 2.01 6.00 .11
> IND,NEAR DEM 6.17 231 2.22 6.14 .15
> INDEPENDENT 5.71 284 2.08 5.72 .12
> IND,NEAR REP 6.41 163 1.99 6.47 .16
>
> NOT STR REPUBLICAN 6.05 337 1.98 6.08 .11
> STRONG REPUBLICAN 6.23 206 2.12 6.38 .15
> Total 6.03 1849 2.09
> 6.05 .05
>
> I apologize for the sloppy formatting, but I'm sure readers of this non-political
> newsgroup are numerate enough to figure out what is going on. Oh yes, P = .008.
May I take this discussion back to some data for a minute? I took the information
provided above, thinking I could do something with it for some introductory students.
Lo! I discovered that it is far less persuasive than p = 0.008 would suggest. If we
compare only strong democrats with strong republicans, the difference in vocabulary
has a
p of 0.045 - usually accepted as significant, but nowhere near 0.008. If I assert
loudly
that the different affiliations can represent an ordinal scale, equally spaced, then I
could do a regression on it. Yes, there is an upward trend as one moves to the 'right'
politically, but it is not all that hot. I'd have to plug numbers a while to work out
the
detailed r^2 value, but it will be small - relieved only by the large sample size. You
really want to stand on that?
Finally, I note that strong Democrats are not the weakest in vocabulary, but
Independents
are. And strong Republicans are not the highest in vocabulary - independents leaning
toward Republicans beat out everyone. Perhaps my claim to ordinal scale for
affiliation
is weak. Perhaps strong Republicans have such rigidity of thought that they don't
learn
new words. (SOrry - so long as we're speculating madly, I had to get in one dig:) ) I
deduce that the p=0.008 was for an AoV analysis. In which case, the alternate
hypothesis
was that _some_ means were different than the others. If we dropped the Independents
and
Ind-leaning Republicans, would we still have a significant difference in vocabulary
skills?
One last point. The average scores for number of words correctly
identified/defined/used
out of 10 runs around 6.0. What was the level of the words used? Can we expect high
school graduates to get higher than 6.0 on this test? Newspapers are usually written
for
an eighth grade level, I'm told. How would eighth graders do on this vocabulary test,
while we're at it?
And yes, you (plural) are right. Vocabulary skills are not required for voting, or
hole
punching, skills. Thankfully.
Jay
--
Jay Warner
Principal Scientist
Warner Consulting, Inc.
4444 North Green Bay Road
Racine, WI 53404-1216
USA
Ph: (262) 634-9100
FAX: (262) 681-1133
email: [EMAIL PROTECTED]
web: http://www.a2q.com
The A2Q Method (tm). What do you want to improve today?
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
