2008年8月20日

CLICK: What makes Olympic medals?

Simple Econometrics model by a few former University student.

Estimated medals Hong Kong and China will get are 3.56 and 91.23. What do you think?

4 則留言:

Sun Bin 說...

interesting research

1) this:

http://c4news.com/livepages/olympics2008/c4/olympicsResults.html

(the motivation -- according to its own confession -- was to find a table so that china is not number one, so it slipped in a 'human right' index) but the GDP and pop indices are interesting.

2) such measures are more useful to show correlation than prediction. so the correct statement is, HK (or PRC) IOC should win so many medals on average in the long term.....instead of the more eye-catching "prediction"

3) there are a few improvement on the model. eg.
a) correlation coeff of log(gdp) and log(pop) are both very low. so the researchers should have tried other way (eg square) because, e.g., for team sport a country need to have a 'critical mass' and a country of 10 people cannot form a soccer team.
b) should show % increase in medal, instead of delta medal for some variables.
c) i think there is not much difference between the soviet and planned factor, could combine as one (if one uses % change then 8 and 11 differes by 27%)

事必大新聞普選網 說...

Doesn't seem to explain why India get nothing.

Josekin 說...

事必大新聞普選網: India has never won big. Notice the "lagged" variable is basically using past performance as a predictor. Here's Freakonomics take:

http://freakonomics.blogs.nytimes.com/2008/08/12/the-gold-medal-chase/

Sun Bin 說...

i prefer the bernard model (per the freakonomic link), sometimes tying all the multi-variables together will blur the real message. bernard's simple, easy to understand and even theorize.

in building these types of models it helps to first theorize and then hypothesize/construct the variable, rather than simpply applying the statistical tools.
e.g. bernard used 'recent hitorical record' to create a continuity between what they were at Athens and the 'theoretical long term average'.

for thoerizing, what i mean is. imagine the host factor for a 1m people country with 2 medal (theoretical or actual), and another with 100m people and 30 medals, does it make sense to add 10 medals to both for the 'host factor'? of course a more reasonable hypothesis is % increase. i.e. the 1m country get (on average) 50% or 1 more medal, while the 100m country will get 50% or 15 medals more. rather than 11 vs 40.

this explains why greece fails the host country test in both models.