Who ruined their family's christmas?

May 25, 2012 12:34 PM

Morbid question about Louis CK's statistical analysis of the crowd at last year's Beacon Theater show.

So in the beginning of his show, he estimates that the crowd was about 2500 people, which was a sufficiently large group to serve as a statistical sample of the US population (roughly). He then goes on to say that, based on a sample of this size, it's fairly likely that 'within two months, at least one of you will die'.

SO! My question is - is it possible to find out if that was a valid prediction? News reports or the like? Googling hasn't turned up anything so far.

posted by FatherDagon to media & arts (28 answers total)

So you're wondering if someone who actually was in that audience at that specific show died within two months of that date? Or are you wondering if the statistic is generally verifiable?
posted by Think_Long at 12:38 PM on May 25, 2012

Audience members of a Louis CK show in New York are probably not a good representative sample group for a bunch of obvious reasons; they probably skew younger, wealthier (those are not cheap tickets), I'd suspect better educated, largely confined to a regional geography, maybe even whiter and male-r than the U.S. as a whole. Most of which suggests a longer live expectancy that a truly random selection from the general population.
posted by 2bucksplus at 12:50 PM on May 25, 2012 [3 favorites]

Specifically if someone who was in that crowd died between then and now.
posted by FatherDagon at 12:50 PM on May 25, 2012

It probably isn't possible to do it except statistically, no. Because the names of the people who attended the show is not public record. Hell, it probably isn't recorded anywhere since even if you had access to credit card records and such there is no way to know if the person who bought the ticket actually attended the show or gave it away/resold it.
posted by Justinian at 1:00 PM on May 25, 2012

No, it's not possible to determine conclusively, because -- barring the chance that it was mentioned in someone's obituary -- you couldn't get a list of all the people at the show.

At best, you could come up with an approximate chance if he was right or not, but your margin for error would be pretty damn high because, as 2bucksplus indicates, you'd start by making guesses at the audience demographics.
posted by Cool Papa Bell at 1:00 PM on May 25, 2012

(The only way it would be possible is if the family of the deceased came forward and announced it, and why would they care about this?)
posted by Justinian at 1:00 PM on May 25, 2012

Your best chance is to comb forums and social media to build an attendee list, and then follow up on those accounts to see if they died.

I think the chances are good someone in that group is dead even though it skews younger. It would be difficult to prove, though.
posted by michaelh at 1:02 PM on May 25, 2012

A better question would be, is he right at all about the basic statement? Given 2,500 random people, one of them will die in 2.5 months. That could be figured out pretty well, just given age and demographic distribution and death rates.
posted by Cool Papa Bell at 1:03 PM on May 25, 2012 [1 favorite]

It's intuitively obvious that he's right if his audience was a representative sample.
posted by Justinian at 1:05 PM on May 25, 2012

It's intuitively obvious that he's right if his audience was a representative sample.

I've seen way too many 'intuitively right' things be totally off-base to trust intuition entirely in any scenario. However, if there were a way to at least ballpark the theoretical likelihood of an audience death (based on a sample that skews younger, metropolitan, slightly above-margin income, etc), that'd be a good place to start!
posted by FatherDagon at 1:12 PM on May 25, 2012

Ok, if you don't trust intuition... The death rate of the population is around 0.08% per month. In a population of 2500, that's 2 people per month. So you'd expect something like 4 deaths out of that population of 2500 if it were a representative sample of the population as a whole.

It's impossible to answer as to the actual audience because we don't know it's actual demographics.
posted by Justinian at 1:17 PM on May 25, 2012 [1 favorite]

But we can say that you'd have to skew pretty heavily to expect no deaths if you expect 4 in a representative sample.
posted by Justinian at 1:18 PM on May 25, 2012

But we can say that you'd have to skew pretty heavily to expect no deaths if you expect 4 in a representative sample.

True, but you wouldn't have to skew heavily at all to experience no deaths if you expect 4 in a representative sample.
posted by milestogo at 1:33 PM on May 25, 2012

Given 2,500 random people

But of course it's not even close to random.

A) They were all healthy enough to be out at a comedy show
and
B) I doubt many elderly people are fans of Louis CK.

There could have been a murder, suicide, or accident, but I think there's also a very good chance all of those people are still alive.
posted by drjimmy11 at 1:33 PM on May 25, 2012 [3 favorites]

The death rate of the population is around 0.08% per month. In a population of 2500, that's 2 people per month.

This is the average death rate over all age ranges. Naturally, it is highest for the most elderly (and at birth, unfortunately). From this link, the death rate for those aged 20-35 is about 1/10 the average rate. If the audience was primarily people in this age range (which seems reasonable to me), then we are down to 0.2-0.5 deaths per month. So while the statistical claim may still be true, I don't think it is obviously correct.
posted by notme at 1:34 PM on May 25, 2012 [2 favorites]

Naturally, it is highest for the most elderly (and at birth, unfortunately)

But even more relevant, as I mentioned above, is that most deaths in first-world countries not at war do not happen suddenly. If they were healthy enough to attend the show only one year ago, the odds go way down.
posted by drjimmy11 at 1:36 PM on May 25, 2012 [1 favorite]

And by "at war" I mean war within the borders of the country. It's also unlikely anyone at that show joined the military and shipped out overseas since then.
posted by drjimmy11 at 1:37 PM on May 25, 2012

The audience could be heavily weighted in the 35ish age range and still contain enough old people to expect a death. In any case I'm dubious that the audience is primarily 20-35. Those tickets are not cheap.
posted by Justinian at 1:39 PM on May 25, 2012

But we can say that you'd have to skew pretty heavily to expect no deaths if you expect 4 in a representative sample.

But as 2bucksplus says, Louis CK's audience is nowhere near a representative sample; it's a convenience sample. The total population includes patients in hospitals and on hospice who have vastly limited life expectancies. For instance, see Table 2.2 (page 24) of the 2009 NYC vital statistics report. The death rate for 30-34 year olds was 0.7/thousand for the entire year. It's 0.5/thousand for non-hispanic whites if we want to assume they made up much of his audience. The rate for his audience would likely be even lower since it's relatively unlikely that someone imminently dying, at a particularly high risk for being killed in gang/drug violence, or such a heavy drug user/alcoholic to cause death in the short-term would be attending the show.

Even so, if we assume the death rate for the audience was .07/thousand and that deaths are randomly distributed throughout the year (not really true, but close enough), that's .00583/thousand/month. Multiply times 2.5 * 2 (for 2,500 people and two months) and you're still only at 0.03 deaths. It is, of course, entirely possible that someone died. Maybe a terminal cancer patient wanted a good laugh in the month he had left to live or maybe an audience member just got hit by a truck walking out of the theater, but our best guess is that it is fairly likely that none of the audience died in the two months after the show.

As a reality check, think about a company with 2,500 employees. If people were kicking the bucket every two months, you would think that something was seriously wrong, and as a SWAG, I'd say that the demographics at an average employer probably skew to higher death rates than at a Louis CK show.
posted by zachlipton at 1:53 PM on May 25, 2012 [1 favorite]

I think you're underestimating how often people die at a company with 2500 employees! It'd be surprising if people weren't kicking the bucket every couple of months.
posted by Justinian at 1:55 PM on May 25, 2012

0.08/thousand was already per month, not per year.
posted by Justinian at 1:58 PM on May 25, 2012

Given 2,500 random people

But of course it's not even close to random.

Yeah, I knew that, if you read the comments.

There's two threads going on here:

* The likelihood of the death among the actual audience.
* The likelihood of the death among an imaginary audience made up of a magically, perfectly random sample from a population.

Whether that imaginary audience is generally American or a worldwide sample is another point.
posted by Cool Papa Bell at 2:12 PM on May 25, 2012

NY Vital Statistics summaries:

http://www.nyc.gov/html/doh/html/vs/vs.shtml

Roughly it looks like for ages 18-45 (summing from the tables) for 2010 in NY the rate was something like 2 per thousand per year.

Given everything described in these tables, I'm actually surprised they don't list number of deaths of people attending a Louis CK show...
posted by NoDef at 2:15 PM on May 25, 2012

In 2009, the age-adjusted death rate for the United States reached a record low of 741.0 per 100,000 population.

In the United States in 2009, 72.3 percent of all deaths occurred among those 65 and older.

Using these numbers, I'm not getting anywhere close to "yes" as the answer to the question, unless you want to argue that Louis CK is super popular amongst retirees.
posted by 23skidoo at 2:27 PM on May 25, 2012

No, it only has to be somewhat less old than the general populace.
posted by Justinian at 2:39 PM on May 25, 2012

Maybe a terminal cancer patient wanted a good laugh in the month he had left to live

I tried to get tickets for that concert and it was already sold out 2 months in advance so any terminal cancer patient would have needed more than 1 month to live.
posted by Obscure Reference at 4:52 PM on May 25, 2012 [1 favorite]

Well, if you're trying to find out if anyone who attended that show died, statistics aren't going to answer that- only the likelihood of it happening. Perhaps you should ask Louis CK on twitter. It seems like if anyone would know it would be him.
posted by oneirodynia at 6:11 PM on May 25, 2012

Also, Louis CK has done a couple "ask me anything"s on reddit. watch for another one and maybe try to raise the subject with him there.
posted by dogwalker at 10:56 AM on May 26, 2012