Misleading language, again

Bob Herbert's op-ed in today's New York Times, Cracks in the Future, is all about the damage being done to California's public universities, Berkeley in particular, by the chaotic state budgeting system. He says (emphasis mine):

More of Berkeley’s undergraduates go on to get Ph.D.’s than those at any other university in the country.

which comes straight off a UC Berkeley website, A Legacy of Excellence.
This is a great example of the sort of language I want my students to learn to look out for. Berkeley's got about 25,000 undergraduates; Harvard's has 8,000. It's not particularly impressive to point out that "more" Berkeley students get Ph.D.'s. Plenty of state schools are larger than Berkeley, but almost every major private university is much smaller.

This is more for someone who teaches physics, as opposed to math and stats, but the New York Times yesterday ran an article about the potential for long-term risks of dementia in high school football players. The article contained the following quote:

At all levels, helmets are safer, rule adjustments have made the game less vicious, and awareness of the hidden dangers of football head trauma — among both medical personnel and the players — has greatly improved. The pressure to perform, however, has never been higher for teams and their players, whose ever-rising bulk and speed collide head-on with immovable physics.

“I go back to Einstein and E = mc2,” said Julian Bailes, a former Pittsburgh Steelers neurosurgeon and one of the leading researchers in the neurological effects of football concussions. “The players are definitely much more massive and that’s one factor. But you have 300-pound linemen running 4.3s — and that factor is squared. The impacts that players face today, not just the big ones that everyone sees but the routine ones in the trenches, is what really worries me.”

It looks like Julian Bailes is mis-remembering the actually relevant equation for kinetic energy=1/2mv², and confusing it with Einstein's famous but enrtirely irrelevant formula.

I do not think that word means what you think it means

According to the UN factsheet on persons with disabilities:

Around 10 per cent of the world’s population, or 650 million people, live with a disability. They are the world’s largest minority.

Huh?

"Exponentially"" actually means something, you know

From an article in today's New York Times about irregularities in the voting for American Idol:

Representatives of AT&T helped fans of Mr. Allen at the two Arkansas events by providing instructions on how to send 10 or more text messages at the press of a single button, known as power texts. Power texts have an exponentially greater effect on voting than do single text messages or calls to the show’s toll-free phone lines.

No, multiplying by 10 is not an exponential operation...

Abuse of percentages

From an article at the New York Times, about the persistent racial gap in high school test scores:

Despite gains that both whites and minorities did make, the overall scores of America’s 17-year-old students, averaged across all groups, were the same as those of teenagers who took the test in the early 1970’s. This was due largely to a shift in demographics; there are now far more lower scoring minorities in relation to whites. In 1971, the proportion of white 17-year-olds who took the reading test was 87 percent, while minorities were 12 percent. Last year, whites had declined to 59 percent while minorities had increased to 40 percent.

I do not think that first percentage means what the journalist thinks it does...

Budget numbers

I've been wondering for a while what Obama meant when he said he was going to raise taxes on families making over $250,000, mostly because that's not a tax-bracket cut off. According to the New York Times summary, His proposal includes undoing the Bush tax cut at the top, apparently creating a new bracket at $250,000 and raising the marginal rate to 36% from 33% for incomes under $370,000, and raising it to 39.6% from 35% for incomes above that.

More interesting to me, though, is his plan to limit the value of tax deductions. I teach my students about the basic structure of the tax plan, and they're always surprised to see the implications. Right now, every $100 that I donate to charity (or spend as interest onmy mortgage) costs me $67, because I get $33 back in taxes. Someone poorer would only get $25 back in taxes; someone richer would get $35 back.

Under the Obama proposal, tax deductions would be capped at 28% (I think this is true even for people currently in the 33% bracket, but under the $250,000 limit). I think that proposal is politically brilliant. "Why should the rich get more help paying their mortgage than the middle class?"

And just in case you thought that budget projections were even remotely credible, the New York Times has a graphic for you:
Obama's projections are, unsurprisingly, on the rosy side, predicting GDP growth in the coming years that's higher than what many economists expect.

Joy of Cooking over time

How do modern home-cooking recipes compare to older ones? Many more calories per serving, according to one study.

Brian Wansink, from Cornell University, directed a study which looked at recipes from the many editions of the Joy of Cooking, first published in 1936. Over the years, eighteen dishes have appeared in every edition, including chili con carne, chicken gumbo, beef stroganoff, macaroni and cheese, brownies, waffles, sugar cookies, and apple pie. The average number of calories per serving went up 63% between 1936 and 2006. (Chili con carne apparently hasn't changed at all, but all the other saw an increase.)

Calories went up for two reasons. First, and most obvious to people eating the food, servings got much larger. The basic waffle recipe used to make 12 waffles; it now makes 6, with the same ingredients. Brownies? Used to make 30, now only 16.

But then there's the caloric density -- more meat instead of beans, more butter, more sugar, so even the same quantity of food has more calories. For fourteen of the 18 dishes, the total recipe has changed significantly, with has an average of 928 more calories, or 44% more. Chicken gumbo now serves 10 people, instead of 14 in 1936, but each serving has 576 calories instead of the old 228. As recently as 1997, the beef stroganoff recipe called for 3 tablespoons of sour cream; the 2006 version calls for a cup. I'd like to take a look at the actual recipes, but as the director of the study says: "That (calorie increases) is more insidious because that's the sort if thing the average person wouldn't notice, wouldn't even think would have happened over the years."

The research was published as a letter in the Annals of Internal Medicine, which means there isn't even an abstract available.

Fuel use per passenger

GOOD magazine has a nice chart answering a questions I've wondered before: what uses more fuel per person, cars or airplanes? (The main chart assumes that a car is only carrying the driver.)
Picked up from FlowingData.

Teens less interested in business careers?

An organization called Junior Achievement, whose goal is to "educate students about workplace readiness," has been polling teens for several years about their ideal job. This year's poll, publicized with the headline "NATIONAL POLL: TEENS' RANKING OF CAREER AS "BUSINESS PERSON" FALLS FROM FIRST TO FIFTH PLACE," showed some major changes from previous years:
"Business person" is down, and "Science/Engineering" and "Doctor" are way up as desired careers. Looks like the economic upheaval has made a huge impact on teens, right?

Not so fast. There's some fine print at the bottom that's important to read:

The 2009 Junior Achievement Kids and Careers Poll was conducted by Opinion Research the week of January 12, 2009, and has a margin of error of plus or minus 3.6 percent. In previous years, the survey's methodology differed, and was conducted using an online survey tool.

I can't find details on their 2007 and 2008 careers polls, but they had many other polls conducted by Harris Interactive, an organization which asks volunteers to agree to be emailed about polls, and then selects a random sample from that pool. (Harris Interactive offers points for agreeing to be surveyed, which you can save towards prizes.) The sample is not remotely random, and the methodology is entirely different -- the poll is online, as opposed to by telephone. Drawing any conclusion by comparing these two older polls with the new random-sample based poll is impossible, because the differences in sample selection and methodology are just too large.

Google insight for search

I've been playing around with the Google Insights for Search tool, and found some really pretty graphs. Here's the popularity of chocolate, as a food and drink related search over time, complete with a big bump each Christmas and a somewhat smaller bump each Valentine's day:

On the dessert theme, we can also see the seasonal rise and fall of cookies, pies, and cakes:

There's a Mason-Dixon line for interest in cookies:
with cake interest centered in the South East:
Finally, Facebook is clearly the new MySpace:

McDonald's or Starbucks?

A recent Pew Research Center Social & Demographic Trends survey included the following question "Just for fun: Would you prefer to live in a place with more McDonald's or more Starbucks?" They've issued a brief report summarizing how people answered the question:
The researchers even carried out a multivariate regression analysis:

It shows that, once one controls for all the factors cited above, the variables that do the most to explain whether someone chooses Starbucks over McDonalds are: having a college degree, having a high income, being a liberal, being a Westerner and being a woman. For example, other factors held constant, a liberal is 13 percentage points more likely than a non-liberal to favor Starbucks over McDonald's.

Or maybe not...

Did you catch the story last weekend about the first woman ever to swim across the Atlantic -- at age 56? It took her 25 days; she rested on a boat each evening and swam during the day. Or at least during most days; she swam on "19 out of 25 days"; "her longest stint in the water was about eight hours, and her shortest was 21 minutes."

Sounds pretty impressive, no? Well, at least it does until you start reality-checking the claim. Swimming 2,500 miles, from the Cape Verde Islands to Trinidad, in 25 days?


View Larger Map

Not possible. A woman swimming 100 meters in one minute -- that's slower than the world record pace for the 100 meter freestyle, but faster than the record for the 800 meter -- would take almost 28 days to swim that distance, swimming around the clock. (And it's entirely impractical for a boat to remain in a fixed position overnight in mid-ocean, when you can't just drop anchor. But that's a minor complication.)

It turns out that she swam only a small fraction of the journey, resting on the group's catamarin for the rest of the trip. The best estimate is that she swam about 250 miles in total -- one tenth of the distance involved in "crossing the Atlantic."

To be fair, apparently Benoit Lecomte, the man who, in 1998, was the first person to "swim across the Atlantic," did much the same thing:

Reached Thursday at his home near Dallas, Lecomte said he did it the same way Figge did — taking frequent breaks on board a boat that moved toward his destination.

He said he doesn't know how many miles he actually swam — "It's a very good question" — and added he, too, took frequent breaks and swam no more than eight hours a day.

Micropayments: the future of newspapers?

Pretty much everyone agrees that the local daily papers are dying, which means that it must be time for someone to bring up the possibility of saving the industry with micropayments. Walter Isaacson, the former managing editor of Time Magazine, writes:

Under a micropayment system, a newspaper might decide to charge a nickel for an article or a dime for that day's full edition or $2 for a month's worth of Web access. Some surfers would balk, but I suspect most would merrily click through if it were cheap and easy enough.

Around the net, people have responded by pointing out that newspapers make their money off ads; subscriptions cover barely more than the cost of getting the physical paper to you in the first play. Michael Kinsley, the founding editor of Slate, provides some telling numbers in the New York Times:

Newspaper readers have never paid for the content (words and photos). What they have paid for is the paper that content is printed on. A week of The Washington Post weighs about eight pounds and costs $1.81 for new subscribers, home-delivered. With newsprint (that’s the paper, not the ink) costing around $750 a metric ton, or 34 cents a pound, Post subscribers are getting almost a dollar’s worth of paper free every week — not to mention the ink, the delivery, etc.

As he points out, if the Times got $2 a month from the one million people who now read the physical newspaper, that would be $24 million a year -- but they take in about a billion dollars a year in ads. To survive, newspapers are going to need to figure out how turn on-line readers into advertising money, not how to get them to pay tiny amounts for their news.

Here are some circulation and advertising figures:
Daily circulation has started to drop off:


And here's the advertising picture, showing that until 2006, the downturn was almost entirely due to the falling market for classified ads, but that in the last year or so, even retail and national advertising has started to drop:

(These numbers come from the Newspaper Association of America; they do not appear to be adjusted for inflation.)

Vaccine impact

It looks like the Andrew Wakefield, the doctor who, in 1998, kicked off the scare over the link between the MMR vaccine and autism faked his data. What effect has this had? In England, before the paper, 92% of kids were vaccinated. Now, it's "below 80%."

While 80% sounds pretty good, you need 95% immunization to be reliably free from measles outbreaks. There were 56 cases of measles in England and Wales in 1998, and 1,348 confirmed cases last year, with two deaths. It had been considered on the verge of elimination in 1994.

Germany had 614 cases in 2006, with two deaths. Those two children were ineligible to receive the vaccine for congenital immunulogical reasons. Such children depend on herd immunity for protection from the disease, as do children aged under 12 months, who normally are too young to receive the vaccine.

The US had 131 cases in 15 states (and DC) in the first half of 2008, with over 90% happening to unvaccinated children. That's the highest rate since 1996.

More economics graphs

Just a quick follow-up on my flurry of economic posts: USA Today has a well-done interactive set of three graphs, done by Moody's Economy, showing different aspects of their job-growth forcast, broken out by job sector:
(link found through FlowingData.)

Unemployment and recession comparisons

There's a nice pair of posts over at Time Magazine's Swampland blog, starting with this graph from Speaker Pelosi's office:
which commits the endlessly recurring error of not taking into account the change in population. I had this on my to-blog queue, but they got enough comments that their economic commentor, Justin Fox, came up with a version that uses the appropriate percentages, instead of raw numbers:

It's still ugly, but no where near as extreme as the original version was. Of course, we still don't know where that pale blue line is going...

Missing the point

CNN's got an interactive map up showing the projected job gain, by state, from Obama's economic stimulus plan. Here's a screen grab; you can roll over the actual map to see exact numbers in each state:
Which states look like they'll be helped the most? California, Texas, Florida, Illinois, New York, Pennsylvania, Ohio, Michigan, New Jersey -- all big states (in terms of population). Why is the plan biased towards big states? It's not. The map is colored based on the number of jobs created in state, but that's not very meaningful. Of course more jobs will be created in states with more people; coloring the states based on the new jobs per capita would be far more interesting.

Or maybe boring. Based on the data from the report, it looks like the states will see very similar job creation rates per capita. This graphs shows the number of jobs forecast to be created per thousand people, ranging from Mississippi, which will get about 11 jobs for every thousand people, to Wyoming, which is forecast to see about 15 new jobs for every 1,000 people. That's not a very large range at all.

XKCD

Every teacher needs a collection of relevant comics, and this one will certainly be showing up in my statistics class in a few weeks, when we talk about boxplots and outliers:

Super Bowl Ad Prices

Super Bowl ad prices provide a lot of opportunities for interesting comparisons and conversations. The original price for a 30 second ad in Sunday's game $3 million, but according to Bloomberg news, only 12 of the 67 spots actually sold for that much, with two of the last ads to sell going for the "bargain" price of $2.4 million (20% off!). NBC had a record $206 million in ad revenue, which would actually work out to an average of over $3 million if there were, indeed, 67 spots. I'm not sure how those numbers relate to one another; the ad revenue total does not include pre- or post-game ads, but maybe it includes non-national spots or "revenue" for those spots promoting other NBC shows?

AdvertisingAge provides a table with a lot of historical context for Super Bowl ad prices from 1967 to 2007, giving the actual price, the inflation adjusted price, the average rating, and the average numbers of homes and viewers watching the Super Bowl. (There's clearly a problem with the listed actual, non-inflation-adjusted, price listed for 1983; it should almost certainly be $400,000, and not $1,400,000. The error doesn't propigate to the inflation-adjusted prices, fortunately.)


MSNBC has a graphic showing the cost and number of viewers over time:

Here's a graph that I created, based on the AdAge data, which shows the CPM, or cost per 1,000 views, in 2007 dollars:
I'd really like to compare the Super Bowl ad prices with the cost of other ads. I'd need to do some hunting to find relevant data.

Finally, and not directly about ad prices at all, the New York Times has a really cool interactive map of popular Twitter words around the country during the Super Bowl (found through Flowing Data.)

Weekly percentages

I'm constantly looking for interesting example for my students that use percentages. I'm going to try to post a round-up of instances I've run in to at the end of each week.

• Tom Daschle is in political hot water right now for failing to pay about $128,000 in back taxes for unreported income in 2005, 2006, and 2007, mostly in the form of a car and driver that he was provided for free. (The amount does not include the interest he owes on the back taxes, which bring the total up to about $140,000.) Given that he's in the top 35% tax bracket, we can calculate the amount that the car and driver cost almost $366,000 dollars over those three years, or about $333 a day.
• A study (nicely described by the New York Times) had 10 and 11 year olds in a "virtual reality environment" decide when it would be safe to cross the street. They did it six times while talking on a cell phone, and six times not talking on the phone. (Half the kids used the phone first, the other half used it after the control crossings.) "[In] the simulation, talking on the phone increased the odds of being hit or almost hit by a virtual car from 8.5 to 12 percent, a 43 percent increase in risk."
  

Robert Reich mixing percentages and raw quantities

In an op-ed in Sunday's Washington Post, Robert Reich wrote:

Meanwhile, our broken health-care system drains more of our dollars yet delivers less care. When President Clinton tried to tackle health care in 1994, it represented 14 percent of our GDP, and 38 million Americans were uninsured. Now, the nation spends 16 percent of its GDP on health, and about 44 million of us are uninsured.

So what's the catch? He's sweeping the population change under the rug. In 1994, the US population was about 260 million; it is now (2008) around 304 million. So the percentage of the population who are uninsured has stayed pretty constant: it's gone from 14.6% in 1994 to 14.5% now. Yes, we are spending more to insure that same percentage, but the numbers he uses in this op-ed exaggerate the magnitude of the increase. I'd call this example intentionally misleading, and not merely confused.