punkwalrus (punkwalrus) wrote,

Lies, Damn Lies, and Statistics

One of the lessons I remember from elementary school was a story about some guy who was a scam artist. It was in a film about how you can make statistics lie. I don't recall who put it out, but it was in some 1970s hygiene film format geared for my age group at the time.

The story was about this detective who wanted to expose someone he thought was scamming potential business owners. He'd loan money to music stores, and make them into a sort of franchise for his company name. Then the business would do poorly, and the owners would need bailed out. "Being kind hearted," he'd buy them out at half their investment, when by all rights, he could legally toss them out on the street with nothing. And this would repeat with someone else.

One set of former owners said that the scammer would supply them with lists of what music people liked. They would poll several thousand people from all over town, and ask them what kind of music they liked and bought. You could even verify the names they provided on the sample. So, based on this untapped market, these franchises would stock a lot of cello music. And very few people bought it. And so they'd be forced to go out of business. Now, most of you may be thinking you know where this is going: the list was rigged, and the sample pollers were plants. They weren't. In fact, they did sample several thousand random people from all over town. And 98% of them said they liked cello music. "Not enough places have a wide selection of cello music."

So, as you can imagine, that's quite a stretch that 98% of the population like cello music. So how did they get those perfectly legitimate figures?

They asked people leaving cello performances at the local music hall.

This lesson is the key to statistics. They can be easily manipulated. They are lies that someone will claim is fact because it involves math. "Aha," you say, my little nerdobites, "what about in a lab? Where they have controls and shit?" No, you have missed the point. Let's take a pretend controlled experiment.

You have three sets of mice exposed to the exact same environment, except one set listens to rock music all the time, the second set listens to classical music all the time, and the third listens to no music. After two weeks, you observe that the behavior of the classical music mice are more relaxed and stress free than no music, and the rock music mice are freaking out and one of them even dies. That's the end of your math right there. This is as far as you can go with this experiment and still stay objective.

But most people would extrapolate that rock music is unhealthy. One of the mice died! And classical music made them all calm. Which would you choose to play for elevator music?


How many thought, even though they may have chosen the opposite just to defy me, "classical music?" Eh? For a second, you probably thought, "classical," and a few of you may have immediately stopped thinking and said, "I bet I am about to be tricked. That Punkie, always setting me up!" (Hopefully you did not say this aloud at work).

Now why would most people choose classical music? Think hard for 1 minute.

BZZZT! I tricked you. Look at that last sentence. "Now why would most people choose classical music?" Hmmm... see it? "Now why would most people choose classical music?" See the subtle trick? I put the suggestion right in the sentence. I suggested, totally pulling out of my ass, "most people." Did I say most people choose anywhere? Did I link to some survey? Show a scientific fact? Even have a lousy footnote or bibliography? No. And that's how I just influenced most of you.

I am chuckling at a few of you who are still arguing in your head about rock mice versus classical mice. Right before the experiment, I said "Let's take a pretend controlled experiment." It never happened. And look how I took that pretend mouse experiment and applied it to humans in elevators. In fact, that whole thing? Completely made up. Now, maybe someone did have that experiment, it sounds like a thing they'd have on Mythbusters. But while this seems like a pure experiment if someone did this, think again.

- What rock music? What classical? There are many types.
- How do we know it was the music, and not say, the frequency of the notes?
- How do we know what volume they played at?
- Why didn't we add bagpipe music? Or country? Or Yodeling?
- Why did we choose mice?
- Why did we do it just once? Why two weeks?

And so on. There's an old joke about experiments and preconceived notions.

A scientist trains a frog to jump at a command. "JUMP!" he screams, and the frog jumps. One day, he cuts off a leg. "JUMP!" he screams, and the frog jumps, but not as far. The next day, he cuts off another leg. "JUMP!" he screams, and the frog kind of does a hop-crawl. The next day, he cuts off the remaining two legs. "JUMP!" he screams, and the frog does not move. "My conclusion of this experiment," he writes in his journal, "is that when you cut off all the legs, the frog goes deaf."

So why this post? Because the media is using them extensively for the upcoming election. Penn and Teller did a bit on it that further explains my point.

Tags: lies, penn and teller, polls, statistics
  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded