Tuesday, October 05, 2004

Talkin' Bivariate Correlation Matrix Blues

From blog The Virtual Stoa, thanks to Carl Wilson for the tip. More on the country music-suicide story below. For me, good music is never depressing, whatever the subject matter. Simply by existing and being good, art is inherently life-affirming.

However, that said, for wonks this is fascinating. Statistics are not my thing though. Surely the correlation between trawling through statistics and suicide is undeniable.

In a follow-up 1994 paper ("'An Achy Breaky Heart' May Not Kill You") published in the same journal, Social Forces, a team of social scientists from Utah State University led by Gary W. Mauk mounted a fierce critique of Stark and Gundlach's findings: problems with ecological inference, they contended, were compounded by a weak causal model. It looked pretty convincing to me, anyway, on a quick skim. Here's the best bit:

Further, while Stack and Gundlach found a relationship between amount of country music airtime and suicide rates... they have not accounted adequately for directionality... Likewise, one cannot determine (1) whether whites who are divorced tend to listen to country music, (2) whether listening to country music tends to cause their noncountry music fan spouses to divorce them, or (3) whether country music makes romantic conflict and divorce seem more normal for those individuals who are contemplating suicide, thus increasing the likelihood that they will attempt suicide...

Reassessing the Link Between Country Music and Suicide
paper from Social Forces journal.

It seems likely any correlation with country music listening is a spin off from more solid predicators such as socio-economic and lifestyle (such as gun ownership) factors. As for making romantic conflict and suicide more normal and attractive, we can hardly single out country music for that.

You know who you are Shakespeare.

Or Charlotte Bronte.

Or the Bible.

No comments: