How not to do statistical analysis

Check this out.  Apparently, an anti-smoking group commissioned a study of whether Minnesota's restaurant smoking ban hurt restaurant and bar business, and the conclusion is that, because employment over 2.5 years was steady, there was no problem.

Now I am no friend of tobacco, and I hate to go places (Lost Wages, Wyoming) where all too many establishments reek of stale cigarette smoke.   Ick.  I'd be glad if a number of these places either closed, or went cigarette-free.

That said, the study involved does not appear to have a control; in general, a 1-5% shift in population is indeed a measurable shift as long as a control is used.  Hence, we do not know whether this is far better, far worse, or about the same as we would have expected without the law's enactment.  Garbage in, garbage out.

It's statistical analysis worthy of the original EPA study used to justify the ban--where the EPA was publicly rebuked by a judge because they had arbitrarily changed the criteria for statistical significance when correlating second hand smoke exposure to the risk of lung cancer.

And what's scarier than the fact that this study apparently made it through planning stages and even peer review; someone probably got their PhD for doing such a transparently flawed study, and may end up teaching your children as a result.