Statistics lie
Statistically speaking, statistics lie. All sorts of people have used all sorts of statistics to reinforce their preconceived notions or advance loony theories. The latest map from DailyKos (via The Corner at National Review Online):
It is unmistakeable what theory Kos is trying to advance. There seems to be a correlation between the locations of the brown dots in the first map to the locations of the blue regions in the second. Ergo, thinks Kos, regions that voted Democrats suffered most of the casualties, the subtext being either that Democrats are more patriotic or that the evil Bush conspired to put more Democrats in harm’s way, or both.
There are several problems with this, in addition to those mentioned in the blog entry at The Corner. The fact that a region voted blue doesn’t mean that the servicemen from that region were all Democrats. Because these blue regions are more highly populated, from these regions even a relatively smaller percentage of people serving in the armed forces would still represent a relatively larger absolute number. To put it in numbers, let’s take two regions, one blue and one red. The blue region has a population of 100,000, with 70% Democrats and 30% Republicans. The red region has a population of 10,000, with 70% Republicans and 30% Democrats. From each region 1% of people serve in the armed forces. Thus we have 1,000 servicemen from the blue region and 100 from the red region. And for the sake of argument, let’s say all of those serving are Republicans. If casualties are evenly distributed, say 1%, there would be 10 casualties from the blue region and 1 from the red. This would still cause the maps to show the correlation between casualties and blueness, despite the assumption that all servicemen from these two regions are Republicans!
Of course, the assumption that all servicemen were Republicans was irrelevent to the actual calculation. What this calculation shows is that the statistical correlation between the two maps shows nothing about the patriotism of Democrats or Republicans.
It is unmistakeable what theory Kos is trying to advance. There seems to be a correlation between the locations of the brown dots in the first map to the locations of the blue regions in the second. Ergo, thinks Kos, regions that voted Democrats suffered most of the casualties, the subtext being either that Democrats are more patriotic or that the evil Bush conspired to put more Democrats in harm’s way, or both.
There are several problems with this, in addition to those mentioned in the blog entry at The Corner. The fact that a region voted blue doesn’t mean that the servicemen from that region were all Democrats. Because these blue regions are more highly populated, from these regions even a relatively smaller percentage of people serving in the armed forces would still represent a relatively larger absolute number. To put it in numbers, let’s take two regions, one blue and one red. The blue region has a population of 100,000, with 70% Democrats and 30% Republicans. The red region has a population of 10,000, with 70% Republicans and 30% Democrats. From each region 1% of people serve in the armed forces. Thus we have 1,000 servicemen from the blue region and 100 from the red region. And for the sake of argument, let’s say all of those serving are Republicans. If casualties are evenly distributed, say 1%, there would be 10 casualties from the blue region and 1 from the red. This would still cause the maps to show the correlation between casualties and blueness, despite the assumption that all servicemen from these two regions are Republicans!
Of course, the assumption that all servicemen were Republicans was irrelevent to the actual calculation. What this calculation shows is that the statistical correlation between the two maps shows nothing about the patriotism of Democrats or Republicans.
posted by Auffie at
12:51
0 Comments:
Post a Comment
Subscribe to Post Comments [Atom]
<< Home