Monday, January 30, 2006

How to lie with statistics, ch. LXXVII

As tallies of this particularly vexing measure pour in from around the land, it's time for another contest. In as few words as possible: What's wrong with this hed? (And, to an extent, with the column, but let's start with the hed.) Usual suspects encouraged to answer. Unusual suspects encouraged to answer. But as you heap vitriol on the hed writer, take a moment to ask your alleged* conscience whether it has any similar sins to confess.

Accuracy improving, but we can do better
A typical Sunday edition of The Indianapolis Star contains close to 480 information elements prepared by our newsroom. These include stories, "breakout" information segments, lists, opinion page commentaries and other items.

In handling that amount of information, errors inevitably occur.

Last year, under our daily "CORRECTIONS & CLARIFICATIONS" heading on Page A2 of the newspaper, we ran corrections for 508 errors.

I take some comfort from knowing that figure is down 15 percent from the 600 errors we corrected in 2004. But the fact that we seem to be improving is little comfort, I'm sure, to the people affected when we sent them to an event at the wrong time, or when we misspelled their name, or when we ran somebody else's photograph above their name.

* OK, if you must: Your arrested-on-suspicion-of conscience. Are you happy now?

4 Comments:

Blogger Dan said...

Well, the cynic in me wonders if accuracy is improving or if they're simply printing less corrections.

4:52 PM, January 30, 2006  
Blogger fev said...

No, no, no. "Cynical" would be noting that an awful lot of corrections seem to be middle initials of really minor bureaucrats, while very few deal with conceptual f***ups, ethical bumbling or the like. Pointing out that 'number of corrections' isn't a valid or reliable way to measure 'accuracy' -- shoot, that's kind of touchingly optimistic.

11:28 PM, January 30, 2006  
Blogger Niko Dugan said...

Or the fact that, using only the grafs you posted, it seems they printed the exact same number of "information elements" in 2004 as in 2005. While

508-600
------- = 15.333%,
600

you shouldn't be "taking comfort that figure is down 15 percent from the 600 errors we corrected in 2004." What if you managed to squeeze in twice the "information elements" that year? They should have been comparing two percentages, not two arbitrary figures.

6:00 PM, February 08, 2006  
Blogger fev said...

Another good statistical point (good to hear from Niko, too). 'Course, if we go much farther down that path, we're going to run into questions like -- well, are all 'information elements' equal? And around the corner from that, are all corrections equal? And that way lies madness ...

11:00 PM, February 09, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home