Reading Between The Lines â Research You May Never Use?

by Philip H.

DISCLAIMER - The opinions expressed in this blog post are those of the author alone. They do NOT represent the official opinion, policy, or action of any governmental agency the author may work for or have ever worked for at the county, state or federal level. If you do not like the content or opinions, contact the author, not your Congressmen.

Sitting in the doctor's office recently I actually got to read a printed copy of the Journal of the American Medical Association. JAMA is both the prestigious place to publish medical research, and something us mere mortals rarely get to actually read. So when I saw that the doctor I was visiting had it out in his waiting room, I snapped it up. Here, I thought, was my chance to see how medical research was published, and thus finally compare it to all the peer reviewed oceanography and fisheries papers I've slogged through over the years.

One thing I noticed was the huge amount of statistics they do, even for double blind controlled studies, and regardless of how large (Or small) the sample size is. This in contrast to how statistics are often used. Something else I noticed is how . . . . interesting . . . . the studies are.

Of all the stuff I read, and the ones I Just skipped, the piece by Redelmeier and Tibshirani stuck with me. Now, if you don't have a subscription you won't get to read the whole thing, so I'll give you the summary. It seems that, due to the increased number of people who are out and about, you are 18% more likely to die in an auto accident on Election Day then other days of the year (Superbowl Sunday included). Yes sir, voting can be hazardous to your health!

Now, I have to admit, had I never seen the issue, and had a really geeky interest in picking it up, I would never have known this. Likewise, I would not have felt the need to seek this statistic out, nor would I have tried to blog it. But I did read it, it did stick with me, and I have blogged it. Ugh

So what does such a number really mean? Googling the web, I found that on 24 September 2008, 18% of voters were undecided; women in Lithuania earn 18% less then men; college financial aid applications are up 18% in Oregon. So perhaps 18% better defines certain parts of our world then we thought. In terms of the election, I do find it interesting that your chance of getting plowed by an errant voter on the way to the polls is just exactly the same as the likelihood that you are undecided in this year's election.

Is it statistically significant in the usual scientific sense? No. Did we, as the general public really need to know this? Again, probably not. Did the doctors involved have a reason other then boredom to look at this question? Yes, I suspect they did. You see, driving is a major source of injury and fatality in the U.S., so it is natural that doctors should be considering why people may or may not be more prone to die in cars. So these two doctors were looking at the relationship between a significant American event, and a significant American activity.

What's my point? Simply this - the next time you hear about a study of driving habits on Election Day, or about methane production on farms, or about why herring spawn on certain seagrass beds in Southeast Alaska, don't be quick to dismiss the study just because you don't see a point to it. True, you as an individual may never NEED to know the facts presented or the conclusions reached. But those conclusions and that study are part of a larger context for the scientists, and they just might enlarge your context as well.

Oh, and buckle up on Election Day - your vote won't count if you don't cast it.

More like this

This is why John McCain is an idiot. He rails against the Montana bear DNA study because he doesn't see the point to it without even considering the larger context of scientific research and the far reaching implications that even the most banal study can have.

Not to be a stickler, but "your chance of getting plowed by an errant voter on the way to the polls is just exactly the same as the likelihood that you are undecided in this year's election." Isn't true. The stat is that there is an 18% rise in accident rate, not an 18% chance of being in an accident.

Some random info about one of the study authors:

Tibshirani is a doctor in the sense that he has a Ph.D., but he isn't an MD. He's actually quite well known for statistics work, having written several books with Trevor Hastie. One of his more recent books is well-known as a textbook for statistical pattern recognition/machine learning/data mining courses.

Given who he is, it'd be odd if there *weren't* a bunch of statistics in the article. But that may not be entirely representative of JAMA publications in general.

Nitpicking noted. I was drwaing a conclusion about two thing that had the number 18% attached to them, but you are correct one is not really the other.

As to the statistics - the Election Day article is actually the JAMA equivalent of a research letter, so its stats are pretty tight. The statistics overload I was referring to is really more focused on the other research articles. They just didn't have as much panache' as this article did.