Friday, January 28, 2011

Scientific Literacy and Copy-Editing, NY Times Blogs Edition

A day or two ago, I happened to comment to friends that people understand science poorly in part because science journalism is often so bad. She's not a science reporter, but this morning, Lisa Belkin had a giant howler on the Motherlode blog:
In the year after an abortion, 15.2 out of 1,000 sought psychiatric help (defined as admission to a hospital or clinic), which was essentially the same as the rate of that group (14.6 percent) in the nine months before the abortion.
I posted a comment about it, because 15.2 out of 1,000 is nothing like 14.6 percent; it's 1.52 percent. I also noted that this was the second time that an error like this had appeared recently on the Motherlode blog. It's just plain embarrassing to see this multiple times. It is just not that difficult to tell what 10% is at a glance, and to tell whether you're looking at 15.2% or 1.52%.

My comment did not get posted, but the text was corrected to the following, without any correction notice:
In the year after an abortion, 15.2 out of 1,000 sought psychiatric help (defined as admission to a hospital or clinic), which was essentially the same as the rate of that group (14.6 per 1,000) in the nine months before the abortion. 

Does the Times have copy-editors? Are all of them innumerate? It's the kind of error that's in the first quotation above that contributes to scientific misunderstanding. And I really think there should be a correction notice on the posting.


calimac said...

I have to assume that the "howler" here consisted of the typographical error of the omission of a "%" sign after the "15.2". Because not only is 15.2% approximately the same as 14.6%, which would cause the second part of the sentence to make sense, if the number is a percentage it is possible to imagine 15.2% of 1,000 women seeking psychiatric help, but if the number is a hard number, it is hard to imagine 15.2 women seeking it. Who is the remaining .2 of a woman? Is she a lilliputian? Did she only get one-fifth of the way through the clinic door?

Lisa Hirsch said...

Yes, 14.6 and 15.2 are very close. The error is an order of magnitude error. 15.2 out of 1,000 is 1.52%, 14.2 out of 1,000 is 1.42%, NOT 14.6%.

14.6% of 1,000 would be 146, which isn't very close to 15.2.

You made the same error Lisa Belkin made in her original blog posting, which I must say surprises me, especially since I give the Times's correction.

calimac said...

The correction doesn't make any sense. There's no such thing as 15.2 women, because there is no such thing as 0.2 of a woman. There is such a thing as 15.2% of 1000 women, however.

Belkin's statement therefore only makes sense if there is a missing "%" after the original "15.2". Then we don't have to deal with imaginary fractional woman, and the rate comparison makes sense.

So I'm not saying what Belkin erroneously originally said, and I'm not making the error you think she did - which was to incorrectly compare hard numbers with percentages. I'm saying they're both percentages, and that Belkin made a typo, not a lapse in reasoning.

Lisa Hirsch said...

I don't know how many population studies you have seen - the researchers always end up with fractions of people. You might not like it, but it's common, and the statistical reasons are well understood by people who do that kind of research.

Belkin's error has nothing to do with fractions of people. There is a gross arithmetical error in the original. She makes an equivalence between 15.2 per thousand (1.52%) and 14.6% of one thousand. That is, between 15.2 and 146. You don't mix hard-number-per-thousand measures with percentages.

It would be correct to say 15.2 and 14.6 per thousand OR to say 1.52% or 1.46%. Mixing the two ways of referring to the same numbers is wrong.

Joe Barron said...

>>There's no such thing as 15.2 women,

Sure there is. If you're talking about a rate per 1,000, you have to include decimals. For instance, 143 stillbirths out of, say, 1.5 million pregnancies would equal a rate of 0.095 per 1,000. There's no such thing as 0.095 stillbirths, but the number is meaningful because it allows you to compare different-sized populations. It's a common calculation. What doesn't make sense is figuring the rate per 1,000 and then reducing that to a percentage. You're creating a needless extra step.