|

 
|
| | Name : | Ryan M. Ferris | Organization : | N/A | Post Date : | 9/30/2005 |
| Comment : | Another statistical normalcy test would be to plot historical
participation rates for every county/precinct and then plot the current election against those.
You could apply the same logic to various independent variables as
accumulated over time:
%percentage voted for incumbent
%percentage voted for Democrat/Republican/Other
%percentage voted for write-in
Other tests could be devised:
Comparison as above between precinct/county as normal with respect to historical comparison. For example, precinct/county blue has voted
Democrat for the last thirty elections with average percentage X. Now
it did this abnormal behavior.... Now we look at every precinct and county
and assign a "normalcy score" and check for large scale deviance...
Obviously, deviance or statistically abnormality just suggests the
probability of some certain event happening. But a suite of tests would
suggest the "statistical abnormality score" of any election. This might
mean a normal statistical correction, or this might be statistically
significant.
We all use this type of logic in our every day lives. If
Hurrican "Stan" develops (tropical wave 99L now forming near Haiti and
the Yucatan Pennisula) to become the third Cat 5 Hurricane in the gulf this
month, we can look at historical records and try to write it off as
statistically abnormal. Or we can look at a 'globally warmed' gulf at
31c 300' down and say, "Houston, you have a real problem...You need to
start believing in a wrathful god, Russian weather modification or that
global warming is real. Your choice." | |
|
|