Estimate how many people who have MeFi accounts are now dead.
December 29, 2004 4:56 PM   Subscribe

Actuary Filter: How many MeFites are dead? [MI]

This is essentially an exercise for me in statistical recognition of how many people are likely to die from a sample that I can relate to (hence Ask not MeTa). MeFi's been going for a while, people who sign up then die would have difficulty in letting us know their accounts aren't useful anymore, statistically how many will be ex-members? Let us know your assumptions.
posted by biffa to Grab Bag (23 answers total)
 
36?

Seeing as the death rate for age group 25-44 in the US is 177.8 per 100,000 population, and that the majority of the 20,000 MeFites are, from my reading, US residents aged 25-44, I would start with that as my first assumption.

Of course, one could actually search out the nations of residence of all MeFites, then average the national death rates by nation proportionally. Then one could try to figure out a way to represent the younger-than-25 and older-than-44 MeFites proportionally. Then one could try to figure out the gender proportions of MeFi and adjust the average according to that.
posted by Sidhedevil at 5:43 PM on December 29, 2004


I was just looking at the 8.34 deaths/1,000 population figure for the U.S. death rate on the CIA World Factbook, when I read Sidhedevil's comment about users from other nations, and that screwed up my potentional calculation.

I was figuring 166.8 out of the ~20,000 users.

.8 because I die a little everytime they make a movie like Dukes of Hazzard.
posted by icontemplate at 5:47 PM on December 29, 2004


Those numbers are per year (I think), which you aren't allowing for. In the 1996 days mefi has been online, and if you assume linear growth, the 20,427 users have been signed up for a total of 55,815 man-years, giving them ample opportunity to get killed.

Using 177.8/100,000 people/year, that works out as 99.2 dead people.
posted by cillit bang at 5:54 PM on December 29, 2004


One problem might be determining how many living-and-breathing (or not) people correspond to the ~20,000 user accounts. Presumably, there are people who have had more than one account over their lifespan.
posted by box at 5:56 PM on December 29, 2004


If we knew the actual number of ex-users, we could work backwards and find out how many accounts were real.
posted by cillit bang at 5:57 PM on December 29, 2004


Oh, yes, per year. I left out the step of aggregating the annual death rate by users by year.

cillit bang, there was a big growth spurt in 2001, so one should aggregate by users per year for accuracy--linear growth isn't really applicable in MeFi's case.
posted by Sidhedevil at 6:16 PM on December 29, 2004


Okay, so here is how I would do this:

I would create a "MeFi user death rate" number based on the proportional national origins of MeFi users and published death rates for age 25-44 for those nations.

Then I would create a "death rate per year" number based on that year's number of users.

Then I would add up all of the years.

Now, that isn't a perfect solution (for example, one needs to subtract the previous year's dead users from each succeeding year's number of users, not to mention the other demographic factors mentioned above).
posted by Sidhedevil at 6:21 PM on December 29, 2004


In the 1996 days mefi has been online, and if you assume linear growth, the 20,427 users have been signed up for a total of 55,815 man-years,

But you're assuming that the 20K users all signed up at the beginning. Instead you have to account for the fact that we started out with only one user, and up until 3 months ago, had somewhere near 17K users. You need to build an equation:

my attempt:

summation of death prob/day *#members on a given day, for every day that mefi has been around.

=death prob/day * summation of #members

assume linear member growth from 1 to 20,427 over 1996 days, so summation of # members = sum(1996,1) of 10.23x.

=.0017*sum(1996,1) of 10.23x dx = sum(1996,1) of .0174x

differentiate and solve to get answer:

= 34.73-.02 = 34.71 dead people give or take.

Or slightly less if growth has been more exponential.

Just my 2 cents.
posted by dness2 at 6:28 PM on December 29, 2004


you're right sidhedevil, I forgot to subtract the dead people from the growth. oh well, they're probably less of an issue than duplicate memberships and all those logins.
posted by dness2 at 6:31 PM on December 29, 2004


Postroad aside, old farts are underrepresented on MetaFilter. Consider that factor.
posted by NortonDC at 9:00 PM on December 29, 2004


old farts are underrepresented on MetaFilter

I'll say!
posted by coelecanth at 10:54 PM on December 29, 2004


I'm shocked that the "I'm dead" and "Me too!" posts haven't started.
posted by bshort at 10:56 PM on December 29, 2004


Great post, Biffa. Great.
posted by Wolof at 11:21 PM on December 29, 2004


Fascinating question -- and to be honest, it had never even occurred to me that any MeFites had passed away (yet).

Maybe User #1 should require some sort of annual "login & test your account" feature.
posted by davidmsc at 12:06 AM on December 30, 2004


Well, what davidmsc says makes me wonder...

How many people haven't logged into metafilter who have an account for the past year. two years? three?

Those would be my first prospects of "bored" or "dead"
posted by filmgeek at 3:31 AM on December 30, 2004


dness2: But you're assuming that the 20K users all signed up at the beginning

No I am not. I assumed linear growth from 0 to 20K by dividing by two, which is a much simpler way to calculate the area of a trapezium.

I don't know why we get such different numbers - where did the 10.23 come from?
posted by cillit bang at 4:37 AM on December 30, 2004


trharlan + NortonDC are right. Metafilter is not a representative sample of the U.S. population or the world population. Definitely not. No infants. Nobody in nursing homes (not that I know of at least, but seriously, if there are 5 members in nursing homes that would be a lot, right?).

So, if we take dness2's WAG at 35, I'd say it's actually much lower. Like half that. 17 dead people are here.
posted by zpousman at 6:32 AM on December 30, 2004


zpousman, we were using the death rates for US people aged 25-44, not for the US population as a whole.

I think my guess of 36 is actually pretty good.
posted by Sidhedevil at 11:18 AM on December 30, 2004


cillit bang,
I realized what you did after I posted, and yeah, I think the only difference is the death rate. I used the 25-44 rate, figuring that probably approximates MetaFilter the best, the youngsters and the old jaded drinkers probably balance each other out.

I think the big unknown is how many of the 20K users actually exist as people and not repeats or junk accounts. So maybe ~20 dead people is a good revised WAG.
posted by dness2 at 12:05 PM on December 30, 2004


Cheers all, glad there are a few others who thought this was of interest - thanks for those who made the effort to answer. The question was actually inspired by looking at number of accounts that are never used (eg look at accounts opened post-9/11 without any posts) plus something that someone mentioned in a post about not knowing what had happened if someone online just disappeared one day. I am quite fascinated by the disconnection of online identity from the corporeal individual and think this is an interesting aspect of that disconnect.
posted by biffa at 3:08 PM on December 30, 2004


A really sad case of "someone online" who "just disappeared one day" can be found here; read the comments on her last entry for the story.
posted by Sidhedevil at 6:27 PM on December 30, 2004


Forty Two.
posted by armoured-ant at 7:32 PM on December 30, 2004


Sidhedevil - could you post a direct link? I can't seem to find the story on her site.

Also, I still miss Turtleguy at Diaryland. He was an awesome blogger that just upped and disappeared. The comments on his last entry ran into the hundreds the last time I looked.
posted by deborah at 10:23 PM on December 30, 2004


« Older Garden Design Software   |   Seances in the San Francisco area? Newer »
This thread is closed to new comments.