With 25,000 towns in the US, is there any surprise that you can find one with an unexpectedly {low,high} incidence of {heart disease,cancer} by pure chance?
This was particularly telling: "... But also leads to stress, ills, and unhappiness that didn't exist in Roseto. Well, until Roseto became just like every other town years later."
Ignoring this, let's look at the structure of this "case/control" setup. The "cases" with unexpectedly low heart disease are all from the same geographic region and environment. Many are related (up to 3 generations living in the same town).
The controls are... well, the rest of America, dispersed across hundreds of millions of acres in diverse environments and with a more diverse genetic background.
In other words, they inadequately controlled this experiment, so any assertions about the results are suspect.
What you say about the experiment results is true. I don't know if this one is any better, but the guy who gave this TED talk found similar results when looking for communities having people with the longest life spans:
One of the places with a lot of centenarians was a place near Okinawa, Japan, where people form life long friendships. They also eat healthy and I'm sure other variables are different. I find the correlations interesting even if they are not conclusive.
This was particularly telling: "... But also leads to stress, ills, and unhappiness that didn't exist in Roseto. Well, until Roseto became just like every other town years later."
Ignoring this, let's look at the structure of this "case/control" setup. The "cases" with unexpectedly low heart disease are all from the same geographic region and environment. Many are related (up to 3 generations living in the same town).
The controls are... well, the rest of America, dispersed across hundreds of millions of acres in diverse environments and with a more diverse genetic background.
In other words, they inadequately controlled this experiment, so any assertions about the results are suspect.