Wednesday, November 21, 2007

Words of One Syllable Dept.

Pause for a moment and think about the number of layers of editing the Stupid Question at right had to go through.

Somebody had to think it was a "By The Numbers" topic. Somebody had to dig up the numbers. Somebody had to set up the display type. Somebody had to come up with "Answered Prayers?" as a hed, and someone else had to approve it. And apparently nobody put on the Magic Science Hat for a second and tried to write these fact claims (this is, after all, the front page of a state capital daily; we'd like to think they're fact-based) out as testable statements.

From June 30 to July 7, Alabama residents were asked to pray for rain. They got it. In August, Birmingham received 4.8 inches of rain, about 1.4 inches more than usual. Totals for September and October were less than normal.

Our independent variable is whether Alabama residents were asked to pray for rain (we're in trouble, you might have noticed, if we try to measure prayer itself), and it has two levels: off and on. The dependent variable is whether prayer is answered, and it's measured by rainfall in a one-month period beginning three weeks after the experimental manipulation (we're hoping the lag is justified in the literature review, about which more in a second).

We're more or less stuck with the one-shot quasi-experimental design, owing to the problems of randomization, control and pretesting. (I mean, the ethical implications of praying for frogs and locusts in Mississippi, or of getting 300 guitarists together to play "Louie, Louie" for rain in Dale, Henry and Houston counties, are pretty obvious.) But even granting the basic "face validity"* of the outcome measure, we still need to quibble about how the DV is operationalized and interpreted. Where's the standard deviation for August rainfall? How are you going to report effect size? How significant are the mean differences for September and October? And let's go ahead and talk about the three-week lag. If that's your way of trying to factor out the confounding factors of death and destruction by tropical weather (just to take one example), it sounds like a bad case of Texas Sharpshooter Fallacy.

Reviewer A, in short, votes to reject. And don't hold your breath waiting on B and C.

As the Georgia experiment has already been discussed here, let's just issue the standard warning and hope somebody pays attention this time. Journalism is about the empirical world. We're not going to run out of observable, measurable things to report any time soon. Please stick to them and leave the supernatural to its own domain.

That's not to say we don't report on the influence of belief on human behavior. We need to keep an eye on people like Bailey Smith who say things like "God does not hear the prayer of the Jew." But we don't do it by running 1A graphics that purport to compare the efficacy of Jewish prayer with Baptist prayer. If you can't tell the difference between stuff you can observe and stuff you can't, now's a good time to consider changing your major.

And next time somebody complains about the secular librul media opening a new front in the War on Xpesmas? Slap 'em across the face with a wet N&O.

* Means exactly what you'd think: On its face, it makes sense to measure rainfall in inches.



Post a Comment

Links to this post:

Create a Link

<< Home