Pages

Saturday, June 11, 2011

The Audacious Epigone

This is not the only site that looks at data and tries to draw conclusions.  Other ones which I discussed were FracTracker on Marcellus Shale fracking and the Incidental Economist which are included on the list of relevant blogs on the right.  Another which I haven't discussed yet but did include on the blog list is five thirty eight by Nate Silver which uses logistic regression to predict the outcome of elections and things like the Oscars.

One which I'm going to discuss now but not include on my blogroll is The Audacious Epigone.  Epigone means a second rate imitator.  Below it's heading it states "Validating stereotypes since 2005."  In the post that I am skewering he or she uses IQ data to argue that McCain voters were more intelligent than Obama's in 2008.   The author places a lot of faith in books like The Bell Curve which argue for racial hierarchy in intelligence and takes data from the general social survey broken down by voting for McCain or Obama and various demographic variables presented below.

Voted for...ObamaMcCainMcCain's edge
All99.9102.52.5
Sex


Men100.4102.21.9
Women99.6102.73.0
Race


White103.2102.9(0.3)
Black92.9n/an/a
Hispanic96.6n/an/a
Asian107.1n/an/a
Party


Republican101.3102.71.4
Democrat99.996.0(3.9)
Independent98.5104.35.8
Age


18-2499.4100.91.6
25-3498.6101.53.0
35-49100.4103.73.3
50-64101.0103.32.3
65+99.2100.51.3

There are obvious things wrong with this reasoning.  The table does look at the underlying minority and gender differences but the author does not look at the possible reasons for these differences such as income, educational opportunities, environmental issues, etc.  All of these variables can interact in complex ways.  Also there is no statistical test reported to determine if any of these differences are statistically significant (in other words outside the statistical margin of error).  This data is based on a sample of 921 Americans which coming from the federal Government is representative of the population but does not include enough minority McCain voters for a comparison (a much larger sample would be needed for that). 

There are also not so obvious conclusions that can be drawn from IQ data.  Sometimes we place such faith in derived statistics from other statistics as if they came to us from God (even atheists sometimes) that we never think to question them when they are really human inventions to explain a phenomena that can't be directly observed.  This includes QB ratings in football, the Dow Jones Averages, and consumer confidence.  IQ and other standardized tests are often criticized for cultural biases in their questions.  Even if the differences were real what do they really mean?  IQ is a composite of different scale scores, a verbal and a performance, which are broken down to still more specific subtests.  An overall difference of 0.3 points between white Obama & McCain voters if real can have an infinite number of possible meanings. 

One final source of differences can be random.  To use a football example, suppose Peyton Manning has a QB rating of 101 next season and his brother Eli has one of 101.1.  That tenth of a point might mean something for fantasy football leagues and their agents might try to make a big deal out of it in contract negotiations but I doubt it matters to their coaches or their dad Archie.

George Bush and Sarah Palin may not be geniuses when it comes to American history but when it comes to pushing people's buttons they do not fall in the idiot or imbecile category. 

**Related Posts**

Probability

The Need for CSI Without Dead Bodies (& Similar Websites)

 

Unemployment: A Universal Underreported Problem

 

Cause & Effect, Slip Slidin' Away