Thursday, October 8, 2009

Case Studies in Bad Statistics

Statistics are vitally important to how we make decisions, and one of
the reasons I believe that every student should take a statistics
class or two is because people often dishonestly or unknowingly
misinterpret the statistics they portray. I present below a few
examples that irked me, just from today's news:

http://www.nytimes.com/2009/10/08/opinion/08kristof.html?ref=opinion

Ensuring everyone has health insurance is a great and important goal
(my issue has always been with the monopsony effect of a public
option, and with finding ways to efficiently fund healthcare, not with
the policy goal of everyone having healthcare), but the statistics and
spin of that article are an embarassment, truly. Because using
improper facts and statistics to justify a cause is inexcusable no
matter how good or bad the cause is, I highlight them:

-It's not 15%, it's 5%, because the other 10% are a combination of
illegal immigrants who don't qualify, people with high incomes who opt
out of employer-subsidized insurance, people already implicitly
covered by Medicare and Medicaid who haven't enrolled yet but would
get benefits if they needed it, and a 6 million person census
overcount. You're improving access for 5% of Americans.

-Bush's tax breaks were 2.4 trillion over 10 years... the terrible
implementation of healthcare that Congress wants would be half of that
*per year*... the comparison is a terrible one. There's also evidence
that Bush's tax breaks actually did help the economy, a lot, for
everyone, but that's a whole separate argument.

-the "socialized health" bit is retarded, because innovation in sewers
and police are less important than healthcare innovation

-there may be 45,000 people who die annually from not having
insurance, but contrary to how he frames it, the net gain in lives
saved would not be 45,000 if universal care were initiated. There are
illegal immigrants, who comprise a very large percentage of that
number, and there are also the people whose lives would have been
saved by innovation that doesnt happen as a result of the bill. On a
net basis, we'd be lucky to achieve a third of that number. Still
significant, but it's gotta be framed properly.

http://online.wsj.com/article/SB10001424052748703298004574455464120581696.html?mod=WSJ_hpp_MIDDLTopStories

- Soft drinks being 5.5% of a person's calorie intake doesn't
acknowledge the fact that quality of calories is very important. 100
calories in vegetables is digested very differently to 100 calories of
cookies; the latter is far more likely to spike your insulin and
result in obesity/diabetes/heart disease etc.

- The fact that West Virginia and Arkansas have soda taxes and have
the highest obesity rates in the nation is completely meaningless. It
is likely that the reason they have soda taxes is because they have
the highest obesity rates in the nation, and it's a way to combat the
problem. From what I understand in WV, the tax, as well as adjusting
the placement and availability of bad food in schools, have been
instrumental in slowing obesity (I haven't seen stats on Arkansas)

http://www.cnn.com/2009/TRAVEL/10/08/flight.delays/index.html
-The argument that "delays have gone up as air travel passengers have
gone up" is somewhat misleading, as well. It could be true, but the
magnitude is much less than implied. The average age of airplanes in
the fleets of the largest airlines has risen over time, and the
variety of planes they fly has gone up as new planes have been
developed but old ones phased out less quickly. Thus, it takes longer
to check all of the planes (mechanics have to look for different
things on every plane, which takes time), when something's wrong it
takes longer to find the correct parts and equipment and use them
properly, and more things go wrong because the planes are older. This
is why "hub" airports, with more big (and old) airlines, have more
delays, while smaller airports, with more small (newer) airlines, have
fewer delays.


http://www.cnn.com/2009/TECH/10/08/social.networks.connected/index.html
-vaccinating just the center of social networks against disease may be
more effective than vaccinating the same number of people selected
randomly, because they may see more people, but unlike obesity, STDs
and politics, a person's charisma presumably has no effect on whether
they get the flu - just their exposure to other people who could get
the virus. If there are 5 friends and one is at the center,
vaccinating the one does not protect the other 4 from giving it to
each other. Technically correct, magnitudal implications far
overstated- the vaccination program certainly wouldn't be "1/3" the
cost. Current methods require about 95% of a population to be
vaccinated for everyone to have "herd immunity"... if this number
reduced it to 85% it'd be miraculous.

No comments:

Post a Comment