Friday, July 20, 2012

where I prove americans are fat using data

The other day, the Economist published a chart in one of its blogs showing a map of inactivity rates by countries. The data is based on a study published in the Lancet, where inactivity is defined as less than 30 minutes of moderate-intense activity five times a week, or 20 minutes of vigorous-intensity activity three times a week, or a combination of both.

Argue all you want with the definition, but I was really curious to see how that correlated to obesity rates by countries. So I looked at it.

I, ahem, borrowed the inactivity data from the Lancet, and got the latest obesity data from the WHO (it's not all super current), where obesity is defined as having a BMI greater than 30. And then I plotted it here:

Of course, that graph isn't very useful, because all of the names collide, and what you really want is something interactive, but I couldn't do that in R very well, so I did it in Highcharts here:

If you click on that you'll be taken to a magical place where you can hover over dots and see countries and relevant statistics. Some countries are missing either due to poor merging on my part or absence in one source or the other.

I also ran a quick and dirty linear model and plotted the line on both graphs (which you can toggle on at the bottom for the Highcharts version). The slope of that line is .2681 and is significant (p value = .01); for every percentage point increase in activity, you'd expect the obesity rate to rise by .2681 percentage points.1 Of course, r-squared on that line is only .096, so inactivity rates only explain 9.6% of the variance in obesity rates; I bet that if we included caloric intakes we'd see a much fuller picture.2

And while this doesn't actually mean anything, the y-intercept of the line is at 6.4; even if inactivity rates were 0, we'd still see a population obesity rate of 6.4. Assuming, of course, that extrapolation of naively-simplistic models based on methodologically questionable data can be trusted. And why not? It worked at LTCM and Washington Mutual and Bear Stearns.

Anyways, the point of this post, like the majority of my complaints, centers around the truthism that Americans are fat. The end. Now off to the gym.3

1. Not percent, percentage points, don't even get me started on that terminology here.
2. Get it? Fuller? Filled out? Because we're talking about obesity?! Because fuller is what you say when someone is fa...okay, I get it, this is not funny.
3. I'd like to thank Highcharts and R for their awesomeness, the Economist for being the best news publication in the history of the universe, and the WHO and the Lancet for not suing me for borrowing data.

6 comments:

brittney said...

I think it's too early to know if they'll sue. But this is awesome, so I'm pretty sure you're okay. ;)

drfindley said...

Way to go Japan. Ruining it for everyone.

But really, I bet if we saw an average daily caloric intake we'd see a much higher correlation.

Or if we took caloric intake and subtracted activity rate * avg calories burned (and some units magic I'm hand waving over), and plotted that against obesity, we'd see some really high correlation.

Also I'm pretending to remember any stats lingo.

jeremiah said...

Were you this interesting before you got married too?

santa said...

Evidence to support the view that some obese people eat little yet gain weight due to a slow metabolism is limited on average obese people have a greater energy expenditure than their thin counterparts due to the energy required to maintain an increased body mass. Thanks.
Regards,
www.creativebioscience.com

Unknown said...

This is disturbing in weight/wait of the up- coming "feast". The American Council on Exercise ACE, says that we consume on average 3000-4500 calories on that day (including 229 g of fat!). Perhaps we should "enlarge" your linear model to exclude the holiday...so that this fatherly model won't be BMIed out? Just saying..

santa said...

Get a birth control prescription. Oral contraceptives are the most common type of treatment for polycystic ovarian disease. The pill helps balance hormone levels and regulates the menstrual cycle. Thus, sufferers enjoy shorter and lighter periods. Thanks.
Regards,
Weight Loss Punch