I can't review this, but perhaps someone on this list can.
please respond to paul.  thanks.
-rich

>Date: Mon, 13 Nov 2000 19:50:17 -0800
>From: "Paul H. Rosenberg" <[log in to unmask]>
>To: Rich Cowan <[log in to unmask]>
>Subject: Statistical Explanation
>
>Rich,
>
>Let me know if this fills the bill as an explanation.
>
>  -- Paul
>
>===========================================
>No Buchanan Stronghold In Palm Beach County: A Statistical Analysis
>Explained
>
>One of the myriad false claims made by the Bush campaign is that Palm
>Beach County is a Bushanan stronghold, and one of the few reasons given
>for this claim is the fact that John McGuire, a Reform Party candidate
>for Congress made a strong showing.  McGuire got 2,651in just a third of
>the county, for 2.1% of the vote--a much better showing than Buchanan,
>who got 0.79%.  It would seem more sensible to call this John McGuire
>stronghold.
>
>And what kind of politician is John McGuire?  Well, for one thing he
>probably wouldn't like me calling him a politician.  I interviewed him
>on Sunday, November 12.  He identified himself with the pre-Buchanan
>Reform Party, and while he wouldn't say anything negative about
>Buchanan, he made it quite clear that he wasn't running on the same
>issues as Buchanan.
>
>But still, the question remains: Does McGuire's relatively strong
>showing provide evidence that Buchanan's relatively strong showing is
>real?  The only way to definitively prove this would be to compare every
>single ballot, and see if the same people voted for both of them.  We
>don't have the data to do that.  But we do have the data to compare
>voting strength at the precinct level.  And here we find that McGuire
>did best where Nader was strongest.
>
>The statistical tool for making this analysis is called a correlation
>coefficient, which ranges from 1.0 (for a perfect correspondence) to 0.0
>(for no correspondence) to -1.0 (for a perfect negative
>correspondence).*  The table below shows that apparent Buchanan
>strongholds were no better for McGuire than Gore strongholds--in fact
>they were just a tiny bit worse.
>
>Precinct-level Correlations Between Reform Party Candidate for Congress
>John McGuire and all presidential candidates.
>
>Nader   0.643
>Browne  0.166
>Bush    0.162
>Gore    0.136
>Buchanan        0.134
>Phillips        0.084
>Hagelin 0.063
>Moorehead       0.035
>Harris  -0.014
>McReynold       -0.063
>
>These figures were generated by a fairly simple process:
>
>(1) Precinct-level data for all Presidential and Congressional
>candidates were converted to percentages.  This was necessary to get rid
>of the spurious correlation caused by all candidates tending to get more
>votes in larger precincts.  This data came from two separate sources,
>the presidential data covered the whole county, the congressional data
>covered slightly less than 30% of it.
>
>(2)  McGuire's data was merged together with the presidential data on a
>single spreadsheet, with all precincts aligned.
>
>(3) The Excel Correl() function was used to generate correlation
>coefficients between McGuire's percentages and those of other
>candidates.
>
>Excel Help explains the Correl function as follows:
>
>Correl
>
>Returns the correlation coefficient of the array1 and array2 cell
>ranges. Use the correlation coefficient to determine the relationship
>between two properties.  For example, you can examine the relationship
>between a location's average temperature and the use of air
>conditioners.
>
>Syntax
>
>CORREL(array1, array2)
>
>Array1 is a cell range of values
>
>Array2 is a second cell range of values
>
>* An internet reference site, The Statistics Glossary
>(http://www.cas.lancs.ac.uk/glossary_v1.1/main.html), explains
>correlation coefficients as follows:
>
>Correlation Coefficient
>
>A correlation coefficient is a number between -1 and 1 which measures
>the degree to which two variables are linearly related. If there is
>perfect linear relationship with positive slope between the two
>variables, we have a correlation coefficient of 1; if there is positive
>correlation, whenever one variable has a high (low) value, so does the
>other. If there is a perfect linear relationship with negative slope
>between the two variables, we have a correlation coefficient of -1; if
>there is negative correlation, whenever one variable has a high (low)
>value, the other has a low (high) value. A correlation coefficient of 0
>means that there is no linear relationship between the variables.
>
>
>--
>Paul Rosenberg
>Reason and Democracy
>[log in to unmask]
>
>"Let's put the information BACK into the information age!"


Organizers' Collaborative      PO Box 400897, Cambridge MA  02140
[log in to unmask]                         www.organizenow.net