Monday, September 23, 2013

Stats in the Real World

Seen today on Instapundit:
MY MENTION OF “WELLNESS PROGRAMS” YESTERDAY produced this email:

At University of Virginia, where I am visiting this term, it’s apparently not just answer-or-pay. It’s sign up your body for Big Insurance’s invasive eyes or pay. Faculty who don’t show up for workplace screening are penalized according to the Aetna health plan “incentives”.
Fundamentally these programs are not about individuals at all. They are about forcing people into being part of massive samples for junk “research” that will hound everyone for years to come. As you know, if you push sample sizes big enough, standard errors fall and you can find practically any point estimate statistically significant, so these huge samples mean ever more “significant” correlations to obsess over…
Anyway please leave my name out of any further discussions. the PC land of campus life could brand me an unwanted guest and I don’t want to go home yet.
...
Meanwhile, here’s more skepticism on wellness programs.
I expect these programs to become more commonplace and more compulsory as government takes over health care.

But I especially like the points made regarding sample size and statistical significance (which is entirely different from practical significance).

Update:  Here's another story that discusses statistical significance, this time in reference to improved performance in a university "flipped" classroom.

No comments: