Books Bygone

Thursday, July 12, 2012

Just your Average American Thinking for Herself

The average American, who should consume about 2,000 calories a day, eats out six times a week. 
The second real sentence in an article by Margo Wootan at The Hill, "Eating in the dark can be dangerous."

Reading sentences like this automatically causes me to go into statistician geek mode. (Haha. That was a statistics joke.) Furthermore, reading sentences like this so early on in an article causes me to read the remainder more critically than I otherwise might have.

Please do read the whole thing yourself. If you are not so inclined, here is a brief summary with commentary followed by a little stats rant (fun stuff).
~~

People eat out more now than in the crappy good ol' days.


Comment: Mind you, there are no studies cited to justify this and other claims, just references to "many studies." Wootan is an expert, though, so take her word for it.

"It’s easy to see why obesity rates are so high — dozens of studies show that eating out more frequently is associated with obesity."

For a variety of reason, restaurant food has a higher caloric content than the consumer may know.

Therefore, "what we eat at restaurants affects our health more than in the past."
To help people make informed choices at restaurants, we worked with Sen. Tom Harkin (D-Iowa) and Rep. Rosa DeLauro (D-Conn.) to pass a national menu-labeling policy in 2010. The law requires calories be listed on menus and menu boards at chain restaurants with 20 or more outlets. Like few policies today, the provision was bipartisan, and it had strong support from not only public health groups, but also the restaurant industry.
Comment: I'm calling chicken s&^t on this. [Links to analyses of proposed legislation of the egg industry.] Sure it had the support of the restaurant industry! If you have more than 20 outlets, you may be able to afford to have every item on your menu analyzed with respect ot caloric content. But which chain is more likely to absorb this expense, the one with 200 outlets or the one with 20?

Comment: I was talking with Mr. Big Food about this. Top's Barbeque came up. Top's has only 17 outlets in the Memphis area. Know what Mr. Big Food said about Top's should they open three more? "There are a lot of Barbeque chains that would love to see Top's go out of business."

Implementation of the legislation has been delayed:
Some conservatives are railing against all regulation, and the administration is being overly cautious as the election approaches.
Comment: Damned Libertarians!
Without clear, easy-to-find calorie labeling, it’s tough to make informed choices for what is a growing and often problematic part of our diets. It’s time for the administration to finalize sensible menu-labeling regulations so people are free to make up their own minds about how many calories they really want to eat when eating out.
Comment: Actually, studies have shown that even verbally telling customers the difference in caloric content of two food choices results in no lasting behavioral changes. If I had a staff of folks working with me, I could cite those studies. Sadly, I do not even have a farm hand.

Comment: Chicken s&^t! [Follow-up to previous link] It has nothing to do with making informed choices blah blah and everything to do with the big corporation-big government  collusion.
~~

The thing is, you can debunk the whole article & its urgent tone if you can show how mis-leading-- if not meaningless-- the second sentence is.
The average American, who should consume about 2,000 calories a day, eats out six times a week.
My Fellow Americans, and the 14 Romanians who read the blog, there is no such thing as an 'Average American.' The terms 'average' and 'mean' describe a characteristic of a sample and of a population, respectively. Both samples and populations are groups of individuals, but these two terms describe the two groups, not the individuals comprising them.
 
REPLY TO LIKELY OBJECTION that this is just "how people talk." Chicken S&^t. 
Margo Wootan is the director of nutrition policy at CSPI. Wootan received her B.S. in nutrition from Cornell University and her doctorate in nutrition from Harvard University's School of Public Health.
I don't have a Ph.D. but I'm quite certain that Wootan knows better-- or should. This, My Fellow Americans, is an example of using stats for your own purposes. And I'm calling chicken s&^t.

For example... .

We may be interested to know, "How often do folks in my little county in rural Mississippi eat out per week?" According to the United States Census Bureau, the estimated population of my little county in 2011 was 10,319 with 75.5% (7791) 18 years or older. Asking each and every one of these folks how often he/she eats out would be cumbersome. So we ask a representative sample-- a room full of, say 1500 folks which looks like the county at large.  (This, by the way, would be a powerfully large sample size. That's another stats joke.)

Note that it's the sample that looks like the entire population-- the sample has the same proportion of whites & black, high school graduates, folks over age 65, and so forth, as the entire population.

We ask each individuals in the sample, "How often do you eat out each week?" We sum the responses, which could range from 0 to 35 (if we assume three meals and two breaks per day), and divide by 1500. Let's say the result is six. If our sample of 1500 folks really did reflect what the entire population looks like in all the ways that might factor in to answering the question-- for example, if there are two restaurant critics in the county, whose job is to eat out, we'd hope (control) we hadn't included either of them in the sample (2:1500 = 14:10,300; there are two, not 14 restaurant critics in the county)-- then we are justified in saying two things:

1. The average number of times folks in our sample eat out is six.

From which we infer

2. The mean number of times folks in the county eat out is 6.

Please note carefully that folks is plural. Just because a lot of people in the sample, and we infer in the population, eat out exactly six times per week doesn't make each of them average and the rest not-normal. 

Furthermore, since we know that there's no such thing as an Average American, it just don't make no sense to make a normative claim about how many calories a nonexistent entity should eat. I'll leave that rant for another day.

Having now thought about that second sentence, six times per week doesn't strike me as all that many, does it you?

6/35 = 0.17

Folks (plural) average eating 17% of their meals out. That means they eat 83% of their meals in.

What's the big deal? It shouldn't be a big deal. It is a bogus deal.

Recall that the range of possible responses to the question is 0 - 35 (three squares and two breaks). Think about what a graph of the of these responses would look like if they were normally distributed-- a classic "Bell-shaped" curve.

X-axis (horizontal) # of times eat out, 0 - 35; Y-axis (vertical) frequency of response
Mean is not the only descriptor of a population. Median is the middle measurement in an ordered set of data: the median of 1 2 3 4 5 is 3; The median of 1 1 1 2 3 is 1.5. It's the 1/2 way point. In the graph above, the median is 18-- midway between 0 and 35.

Mode is the most frequently occurring measurement in a set of data. In the graph above, the Y-axis is the frequency of the response-- how many folks said 0, 1, 2 ... 35. The most frequent response is 18.

In a normal distribution, the median, mode, and mean are all the same.

It should be intuitive that if possible answers to "How many times do you eat out per week" are anything like "none," "three," "breakfast, lunch and supper ever weekday," than with an average of six, the responses can not be normally distributed.

Same axes as above

Plotting the frequency of the response against number of times you eat out per week would look something like this-- except even more skewed to the left.

The median is still 18. However, the response that was given most frequently-- the mode-- is less than 18. And look what happens to the mean!

To make things simple, let's suppose we only sampled 150 people. And let's suppose that they all said they eat out six times per week. We should question our sampling techniques, but let's suppose they were fine. So six it is. But let's suppose 148 eat out six times and the two restaurant critics each eat out 35 times per week. What's the average now? 6.4 What's the big deal?

Let's suppose we were really sloppy and only sampled 15 folks-- 13 and the two critics. Average number of times folks eat out each week? 9.6

The point is, as everyone who's taken Intro to Stats knows, the mean is not robust to outliers.  Outliers pull the mean in one direction or the other.

What's the big deal? The big deal is that in an asymmetrical distribution-- one skewed left or right-- the majority of folks' behaviors (eating out) are not well-described by the average/mean.

Unless you think the situation through, as we just did, you will have the same "Holy Cow! Are you kidding me? People eat out six times a week? Wow. That's a lot" initial reaction to that second sentence as I did. That's how Wootan wants you to react.

We know better. :-)

By the way-- if you are interested in learning the nutritional value of foods you cook at home, Nutrient Facts is a great place to start. You enter in a food, or the ingredients for a recipe and it spits out the facts.

2 comments:

  1. It also sounds like lunch every day of a work week, plus one night out per week...

    It seems like not very many people "brown bag" it these days...

    ReplyDelete

Be nice. Nothing inappropriate, please.