study: 1 in 12 legally drunk when leaving MLB or NFL game

Faced with watching that outfit on a consistent basis, drinking to oblivion is preferable.

 

You've got a point with the potential for selection bias, but a sample of 362 is actually pretty powerful. At that sample size, and given that point estimate, you would be 95% confident that the interval between 5.49% and 11.18% contains the true proportion of people who meet the legal standard for impairment. If you double that sample size, your interval only shrinks to 6.32%-10.35%. If you make it 1,000, your interval's 6.62% to 10.05%. So as far as drawing inferences about the drive-worthiness of departing MLB/NFL fans, the larger samples don't seem, to me, to convey sufficiently better information to justify their cost (i.e., effort).

 

I'm surprised that many people can afford to get drunk at games anymore. I don't think I've had more than two at a game in about five years.

 

Smuggling in a few tallboys is a lost art.

 

That is using the typical formula. Problem is this isn't a typical study. Just like online studies are wildly believed to be next to worthless because it is a biased pool of people, this too has a bias in the pool of people. Not everyone has internet and many who do don't do surveys just as a lot of people would never stop and do a breathalyzer outside a sports stadium.

That is only half the story though. Doing 13 of the 30 stadiums is a joke. It is the same thing as a poll for a national election only talking to people from 20 of the 50 states. Every group must have representation for any survey to be credible. The NFL part is much, much worse. Three out of 32 stadiums. Seriously? Less than 10 percent even had a chance to participate and you want to use it? Simply horrible.

 

The "typical formula"? Is there some other formula for the 95% confidence interval for the population proportion? I was addressing your point regarding the sample size and pointing out that the additional information gained might not be worth the cost incurred in gaining it.

Re: the selection bias. Yes, self-select is a potentially serious issue here. But unless you have some solid expectation that the stadiums selected are not representative of the typical NFL/MLB stadium with respect to this issue, then your complaint about sampling only a portion of the stadiums is without merit. And it is NOT the case that "every group must have representation for any survey to be credible."

 

Hmmm ... "doctorquant" or "Cubbiebum" on the matter of sample sizes and statistical significance ... hmmm, the college professor or the, um, not college professor ... boy this is a toughie

 

See, i was thinking media was the 1 in 12.

 

Mother is a sociologist and I worked for years in the Purdue research lab conducting the surveys. Even amongst sociologists there are big debates of the credibility of how studies were done. This one, from my experience, would be laughed at by most sociologists.

And yes you do have to give every group a chance to be included. In national polls you don't have to make sure every race is represented but you have to give them the chance. You don't call only wealthy neighborhoods or only call poor neighborhoods nor do you call only neighborhoods with a very high percentage of white people or visa-versa.

The stadiums, cities and areas can be very different. Milwaukee doesn't have a whole lot in common with LA. Heck the Cubs don't have much in common with the White Sox in terms of the type of fans that come to both stadiums. You have to get every city in the poll or you lose a lot of creditability. The study, while interesting, proves practically nothing because of all the problems with how it was done.

I don't know everything about sociology and studies but I do know quite a bit from being involved in a lot of them.

 

You are totally correct, but that doesn't necessarily mean that both Milwaukee and LA have to be included in the survey for it to be valid. The question is, to what degree are Milwaukee and LA different with regard to attendees' drinking patterns? If there is no reason to think (or evidence to suggest) that the typical Brewers fan drinks more (or less) than the typical Dodgers/Angels fan (or that one fan or the other drinks more or less than the typical MLB attendee), then there's no need to dismiss the study simply because one or the other city was not part of the sample.

If the authors' survey was dominated by eastern, urban stadiums -- e.g., Fenway Park or Yankees Stadium -- I might be inclined to think their results overstate the impairment rate. This would be because, as I understand it, the typical attendee there gets there by using mass transit. Therefore, it would be reasonable to think that there is a greater proportion of impaired attendees simply because a smaller proportion of them are driving. The converse would also be true; if these types of stadiums were underrepresented, then perhaps the overall impairment rate is actually understated by this study.

In the social sciences milieu, a study's validity (particularly along these lines) is almost always an open question, and I have had my share of go-'rounds with journal reviewers (and with authors when I am serving as a reviewer). The question is never answered definitively -- how could it be? -- but one tries to cover one's rear end in as many ways as possible. Would this study be "better" if more stadiums had been sampled (or more people surveyed)? Of course. Would it have been more informative? That answer's a whole lot less obvious.