Block, SkipCaptcha, Automoderated users, edit
752
edits
Changes
Talk:2012 Summer Olympics (view source)
Revision as of 20:41, August 13, 2012
, 20:41, August 13, 2012→Will countries that have implemented same-sex marriage underperform?
: 4% isn't a significant difference anyway, unless it continues. One sample showing a 4% change doesn't really prove anything by itself. You could be right, but we'd have to analyze it long-term, rather than just one year, given the small change. [[User:Gregkochuconn|Gregkochuconn]] 16:14, 13 August 2012 (EDT)
: 4% isn't a significant difference anyway, unless it continues. One sample showing a 4% change doesn't really prove anything by itself. You could be right, but we'd have to analyze it long-term, rather than just one year, given the small change. [[User:Gregkochuconn|Gregkochuconn]] 16:14, 13 August 2012 (EDT)
: Assuming GregG calculated the p-value correctly, he is correct in that there is no statistical significance just from this one test. Generally, you would need a p value of 0.05 or less for it to be significant (or by stricter standards, 0.01). 0.73 is nowhere near that. I'm not going to verify the calculations, but it seems right. And assuming it's right, GregG is correct that it is not significant. However, if repeated tests showed a 4% decrease, that would be significant. We can run it again in 2014 and 2016, and beyond. And possibly we would have more nations to work with in those years. Keep in mind that the Winter Olympics (2014) have fewer total medals, so a 4% drop would be less significant, by itself. But if demonstrated, it could show significance of the trend. Something inonsistent with the 4% decrease, such as a small increase, would indicate no correlation. Bottom line is that this one sample test can't disprove the correlation by itself, but it doesn't prove it either. I'm an actuarial science major at school, so I'm pretty knowledgeable about this stuff. A repeated 4% decrease would be significant. But for the moment, it's not. I'm going to retroactively analyze the 2008 medal data, although we'll have fewer countries. When I'm done, I'll post the results. [[User:Gregkochuconn|Gregkochuconn]] 16:26, 13 August 2012 (EDT)
: Assuming GregG calculated the p-value correctly, he is correct in that there is no statistical significance just from this one test. Generally, you would need a p value of 0.05 or less for it to be significant (or by stricter standards, 0.01). 0.73 is nowhere near that. I'm not going to verify the calculations, but it seems right. And assuming it's right, GregG is correct that it is not significant. However, if repeated tests showed a 4% decrease, that would be significant. We can run it again in 2014 and 2016, and beyond. And possibly we would have more nations to work with in those years. Keep in mind that the Winter Olympics (2014) have fewer total medals, so a 4% drop would be less significant, by itself. But if demonstrated, it could show significance of the trend. Something inonsistent with the 4% decrease, such as a small increase, would indicate no correlation. Bottom line is that this one sample test can't disprove the correlation by itself, but it doesn't prove it either. I'm an actuarial science major at school, so I'm pretty knowledgeable about this stuff. A repeated 4% decrease would be significant. But for the moment, it's not. I'm going to retroactively analyze the 2008 medal data, although we'll have fewer countries. When I'm done, I'll post the results. [[User:Gregkochuconn|Gregkochuconn]] 16:26, 13 August 2012 (EDT)
: Here is the medal data from previous years from the original post, omitting any countries that legalized same-sex marriage after 2008
Belgium - 5 in 2000, 2 in 2008
Canada - 12 in 2004, 18 in 2008
Netherlands - 25 in 2000, 16 in 2008
Norway - 6 in 2004, 9 in 2012
South Africa - 6 in 2004, 1 in 2008
Spain - 19 in 2004, 18 in 2012
Total - 73 expected, 64 actual
This is a decrease of 9 medals out of 73, a 12.3% decrease.
Let's look at the countries that legalized same-sex marriage between 2000 and 2004, as far back as we can go.
Belgium - 5 in 2000, 3 in 2004
Netherlands - 25 in 2000, 22 in 2004
Total - 30 in 2000, 25 in 2004
This is a 16.7% decrease, but with far too small a sample size to be significant by itself. [[User:GregG]], care to run the significance test?
Also of note, there appears to be something going on with the Belgians and Dutch, who have seen a consistent decline, while the rest of the nations have not, apparently. Since they are geographically and culturally close, this may be relevant. Ignoring them, we have no data for 2004, a 4.6% INCREASE in 2008, and again a 4.6% increase in 2012, neither of which is statistically significant by itself, and probably still not together. (Note: 4.6% increase is just as statistically significant as 4.6% decrease. In this case, not very) Of course, we can't just kick the two lowest samples out of the data set. But there seems to be some factor that has steadily brought the Dutch and Belgian totals down, without influencing anyone else. Was something else going on in Belgium and the Netherlands at the same time? Or was the 2000 medals we're using as the baseline simply an anomaly? I'm guessing the latter. 2000 was the Dutch's best Olympics EVER. It wasn't the Belgians', but even the 40% decrease from 2000 to 2004 isn't significant when they only had 5 to begin with. I'm going to rerun this data when I have more time, using totals from the three previous summer Olympics, and totals from all Olympiads thereafter, multiplying the latter by 3 or 1.5 if necessary, where there are fewer years, although I'll note where so as not to effect the significance test. [[User:Gregkochuconn|Gregkochuconn]] 16:41, 13 August 2012 (EDT)
== Medal weighting ==
== Medal weighting ==