We listened to an interesting podcast about numbers while driving home from Thanksgiving dinner with the wonderful sons. The interviewee is a journalism professor who's just published a book about what he calls "the dark side of mathematics."
Most of what he dealt with was the statistics branch of math. No better way to lie to people than by using statistics.
The amusing part of the podcast came when the prof failed to properly apply his own principles.
Citing a study that showed men have an average of seven sexual partners in their lifetimes while women report only having four, the prof declared that this was impossible, that the numbers should be close to even.
But he forgot that the study involves sampling a population and making generalizations about the population. Of course, this is the basis of modern polling, and it works in various levels of success.
But consider this. We have an unusual sample of eight people, one guy and seven women. Now, it happens that the guy happened to sleep with all seven women. So he reports having seven partners. Now what if each woman also slept with three other men during their lifetimes. Men not represented in our sample.
Bingo. We've matched the study results.
This is, of course, wildly simplified. But in a study like this one, unless you've a huge sample, you face a similar difficulty. The results may be absolutely accurate and still not be representative of the population as a whole.
But in one of those delicious coincidences often mistakenly referred to as irony, the next morning a news show I was watching reported on a "study" conducted by Popular Mechanics (one of my favorite sources for reliable research!) that showed the U.S. Postal Service handled packages more gently than FedEx and UPS.
They even included graphs, another technique the professor noted as being useful for deception, and one of my personal favorites. By manipulating the scale, you can make relatively small differences seem much greater.
Since the title is about three kinds of lies, I'll mention one more -- even though that's not what the quote refers to.
The professor noted that many studies the news media report on show correlations. You know, people who eat kale tend to have fewer cancers than people who don't, or people who drive red Fords have fewer accidents than people who drive black Pontiacs.
But correlations don't necessarily indicate cause. They can help point the way, but correlations are always starting points.
I learned so much about how to properly use numbers, I decided to do a little research and report on the results here.
Consider the graph above. I took a representative sample of approximate IQ scores of people I know read these ramblings. The first thing to notice from this graph is that I am dumber than my readers.
Note, however, that I am not myself particularly dumb. The graph clearly shows I am way above average.
But the obvious conclusion to be reached here is that reading my blog makes you smarter, so I encourage you, dear reader, to pass this along to all your friends.
