After finishing my Physics degree I took a year studing pedagogics, and read several papers using regression analysis to find correlation between sociological parameters like grades and family income. After seeing what you can come up with I have lost my faith in regression analysis as a scientific tool unless its used in conjuction with some repeated experiment.
This is a classical example, note in specific the line:
"But even after controlling for nearly all imaginable variables - socioeconomic status, level of physical activity, number of close friends, quality of social support and so on"
Since you only can account for almost all variables but not all I do have trouble believing it. The important part is the causal explanation offered. Is it sound?
In this case it actually might be a sound explenation:
One important reason is that alcohol lubricates so many social interactions, and social interactions are vital for maintaining mental and physical health. As I pointed out
last year, nondrinkers show greater signs of depression than those who allow themselves to join the party.
But actually most people I know who abstains are more social than my average friend, so I don't think this will apply for them anyway