Sunday, April 03, 2005

Statisticsbusters

I'm quite a fan of the show Mythbusters on the Discovery channel. It's fun to watch them light things on fire, drop things from heights, and blow things up in the name of science. And it's refreshing to see science being used in order to debunk some obviously false myths.

I was less than impressed with their conclusions from tonight's experiment to see whether yawning is contagious, though. The setup: get 50 volunteers, dividing them into two groups. One-third are placed individually in isolation as a control; the other two-thirds receive the same treatment, but the greeter first yawns at them. 25% (I think) of the control group yawned in the allotted time, compared to 29% of the experimental group. The average yawn time was 9.6 minutes in the control group versus 9 minutes in the experimental group. From this the mythbusters concluded that the hypothesis is confirmed. I argue that the test is inconclusive, since the result is not even 1-sigma.

Of course, the mythbusters often have to work in suboptimal conditions with an insufficient sample size to test their myths, and in many cases one can legitimately argue with their methods. Since it's all for entertainment purposes (and statistics is hardly entertaining), all of this is excusable. Still, it's unfortunate that a show that advances the scientific method would so cavalierly brush aside statistical considerations.

6 Comments:

Blogger bdean said...

Hmm... I started watching that episode, but fell asleep halfway through :)

4/04/2005 11:58:00 AM  
Blogger Ben said...

In an amazing coincidence, this was one of the few episodes of the show I've ever watched. I agree completely on the lack of statistical evidence. If I recall, one of the other tests they did was with buttered bread (to see if it lands butter side up or down). I think they could have done a better job of explaining the physics of the situation, since the counter height has a great deal to do with how many rotations the bread undergoes before it hits the ground. I guess they didn't want to bore the watchers with the *yawn* boring details.

4/04/2005 04:38:00 PM  
Blogger acg said...

I remember once reading an article that explained why your toast always lands jam side down. It went further and explained why this would happen even if you were an alien on an alien planet; your (alien) toast would still land jam side down (because the counter height would be related to your height, which would be related to the gravity of the planet in question). The universal law of dropped toast: it always lands jam side down. :)-

4/04/2005 11:31:00 PM  
Blogger Justin said...

I *loved* that article, though I've long since lost the link... Anyone happen to have it bookmarked?

4/05/2005 04:14:00 AM  
Blogger Eric said...

I think you should write them. Perhaps they could do an episode where one of the myths involves how people respond to some question. Then an important part of the episode would be how many people do they need to ask, and is the difference significant. Of course, that's not nearly as good as bringing careful statistical analysis to each of their subsequent shows, but perhaps it's a start.

But it's hard to blame them. I still feel like I'm pulling teeth with PhD astrophysicists who seem to have trouble understanding basic statistical concept.

4/06/2005 04:53:00 AM  
Blogger Justin said...

If you think that's a pain, try talking statistics with computer programmers and/or business folks.

4/06/2005 11:01:00 PM  

Post a Comment

<< Home