Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
The oldest person in the world never holds that title for very long (fivethirtyeight.com)
56 points by Tenoke on May 26, 2015 | hide | past | favorite | 52 comments


I invented a puzzle some time ago I haven't solved yet:

Calculate how often on average oldest person in the world dies.

Assume whatever is necessary to give the answer. For example that population pyramid is constant over time and known.



Awesome. But forgive me not reading the answers. I want this to remain puzzle for me. I don't want to learn the answer. I want to invent it some day. Same way I feel about Rubik's cube.

I still have fond memories about finding out precise recursive equation, implementing it in LOGO for Atari and waiting minutes to find out the answer to how many times you have to try to burn all matches in a box by picking them from the box one by one at random, lighting them and putting them back in the box before picking next one.


Oh, neat - I wrote the top answer there (over two years ago apparently!)

I stand by the modelling, although it's embarrassing to read my two year old code.


I found that model somewhat overkill. I would have cut some corners there: assume that the oldest person alive is 110 or older (reasonable, given a world population in the billions and the data we have) and compute his (more likely her) hazard rate from the curve you fitted.

For hazard rates at old age, taking life expectancy equal to 0.5/hazard rate isn't an extremely bad approximation (I have no education in actuarial science and gave that little thought; corrections welcome), so one gets:

   age   h(age)  life expectancy
   109   0.473   1.058
   110   0.503   0.993
   111   0.536   0.933
   112   0.570   0.876
   113   0.607   0.824
   114   0.645   0.775
   115   0.685   0.730
   116   0.728   0.687
Again, we get in the range of that 2/3-ish.


Reminds me of the guy who answered some question about Python how to put two strings together, and then 3 years later commented 'Oh look here I am asking the same question I answered' lol


What i find interesting there is that multiple approaches seemed to arrive at about the same result, 2/3 of a year between deaths. That makes me think it's highly probable that it's the right answer for the current time period at least.


Note that one of the answers there is based on the GRG data used in OP. Different graph but same conclusion - average tenure has dropped, Calment was a monster.


Saying "one of the answers" is a very modest way of referring to your own answer ;)


It's just a Poisson process


...Except that:

* Birth rates are changing

* Mortality rates are changing

* For any one person, the chance of dying (generally) increases as they grow older

* Deaths are correlated. For instance, things like the 2003 tsunami.


I could be satisfied by solution that doesn't take into account three of your points. But the 3rd one is something I can't ignore.


"73-year-old granddaughter". That is just mind-blowing.


yeah, nuts. I wonder how many generations down she has met.


I can't stand this parenthetical. it's not strange at all!

(Strangely, the 20th century is considered to have started Jan. 1, 1901, so there are a handful of other, verified women still alive who were also born in the 19th century.)


Yes it is. You're confusing what's right with what seems intuitively wrong to people. It's strange because it's inconsistent with most systems which group numbers by significant digits. For example, a person who is 30 years old has entered their 30's.

For someone who is clearly perturbed that most people don't know this archaic and inconsistent rule, I'm surprised you're perturbed by a writer who is helping your cause by including a parenthetical. I suspect the writer selected the word "Strangely" to prompt the reader to pay special attention to the rule, because what he's about to say is going to violate many people's intuition and assumptions.

If you can't stand that people are unfamiliar with this rule, then this writer should be on your Christmas list.


One of my favorite things about HN is when people take the time to be extremely clear about something that's almost obvious. Always gives me a chuckle.


> For example, a person who is 30 years old has entered their 30's.

This is a different linguistic construct. We have decades like the 1990's, which are like the 30's.

Linguistically, we have:

1800's = 1800-1899

19th century = 1801-1900 (since AD started in 1, so the first 100 years, or century were 1-100)

One of those forms is linguistically similar to your 30's, the other is not.


Okay, but you enter your fourth decade on your 30th birthday, so there's less counterintuition involved in the age realm.


> inconsistent rule

Imho, it is not inconsistent and it is not a rule. It is a fact (as long as one uses the Gregorian calendar where there is no year zero[1]).

[1] http://en.wikipedia.org/wiki/0_%28year%29


This relates to the argument about whether array indexes should start at one or zero. The standard argument for one-indexing is that it is what people do. But do they really do that?

The first century is 1 to 100. The second century is 101 to 200. and so on. After a while things get a bit blurry. Eventually the 21st century is 2000 to 2099, as proven by humans holding their big parties on Jan 1st 2000 instead of Jan 1st 2001.

Humans start at one, but weary of the strict logic of it. Without acknowledging it and without agreeing which century, humans sneak in a 99 year century so that centuries work as if zero-indexed; 0-99, 100-199,...,1900-1999, 2000-2099.

What then of the argument "Array indexes should start at 1 because that is what humans do."? It fails two ways. First, that is not in fact what humans do, (and the parenthetical strangely is evidence of that.) Second, humans weary of strict logic. Having the computer do what humans do is a bad idea.


Maybe the big parties were not so much to celebrate the arrival of the 21st century, but the arrival of the magical year 2000 with its flying cars and stuff.


>> (Strangely, the 20th century is considered to have started Jan. 1, 1901 > it's not strange at all!

Are you a Lua programmer?


Wow, is Lua now the canonical example of a language that uses 1-origin indexing? No one remembers Fortran anymore? God I feel old...


Actually I just went here: http://en.wikipedia.org/wiki/Comparison_of_programming_langu...

sorted table by "Default base index" and at first picked FORTRAN, considered COBOL and Smalltalk for a while then saw Lua and thought that should resonate with people. My personal experience with 1 based indexing is from Matlab and Pascal (though you can base your indexes on whatever value you want there).


One of the few that people still care about!

Aside from BASIC, maybe.


Or Algol. Or Pascal. Or Smalltalk...


There is no Year 0. So the first century was Year 1 - Year 100, the second century was Year 101-200 etc.


"Year 1" was originally conceived as "the first year of our lord", which is one reason why there's no "zeroth year ...".

Now we're in the "two thousand and fifteenth year of our lord". Imagine all the off-by-one errors if years were zero-indexed.

Also, it wasn't until Fibonacci in 1200 or so before the western world developed a working relationship with zero.

But natural counting numbers are always 1-indexed anyway. The second dozen starts at 13. Bakers excepted, of course.


But measurement of a person's age is 0 indexed. It is a year after you are born that you are 1 year old.


Sure, but while you are 0, you are in your first year of life.


Natural counting isn't 1-indexed. You usually count things after you tally them, not before. (You don't start counting beans by saying "one bean" and then waiting for that bean to arrive.)

Besides, it is not all that clear if someone just says "Jan 1 marks year one" whether it marks the end of the first year or the beginning. With dates, it usually marks the beginning. That's why it is 1-indexed. We hardly count anything else that way, so why would you assume it is 'natural counting'?


So is Wikipedia (http://en.wikipedia.org/wiki/20th_century) wrong on this one? Seems like all other centuries are calculated correctly.


Yes it's wrong. According to that page and the one on the 21st century, the year 2000 didn't belong to any century. Oddly, the wrong statement is listed as having two citations.



As of right now it appears to be correct (1901-2000, inclusive).


Out of curiosity, what time interval do you think of when you hear the term "the 1900s"?


(Not couchand, but) 1900-1999, which is not the same thing as the 20th century.


Yeah, not sure what the "strangely" is doing there.


Yeah, sloppy.

It's sad that 538 has turned into a pop-stats site that runs on volume and snappy headlines.


>It's sad that 538 has turned into a pop-stats site that runs on volume and snappy headlines.

It's our fault, they're catering to our demands. You can't run a media company any other way. No one wants to read read (although I'd bet on average everyone spends far more time reading). So you get wham-bam, buzzfeed-esque content. Even on HN, long form articles will inevitably be met in the comment section with a "tl;dr?".


There's nothing wrong with asking for a tl;dr when you're trying to find out if the long form is worth your time. (Most are not.) In academics, the tl;dr is called the abstract, and it would be silly to think that they should be eliminated because people ought to just read everything.


The real problem seems to be that Nate Silver's (and 538's) brand value exceeded the available payment from NYTimes. So he did the rational thing and tried to capture it in the open market.

But Nate's a stats guy, not a marketer. Brand dilution is a thing. I don't begrudge him for selling out. I just wish my domain-based trust filter hadn't been broken in the process.

I sure hope they put together some solid election coverage.


> But Nate's a stats guy, not a marketer. Brand dilution is a thing. I don't begrudge him for selling out.

Incidentally, Nate didn't sell out when 538 left the NYT, he sold out when 538 was picked up by the NYT.


Monetarily, perhaps -- which if true makes the subsequent devaluing all the more disappointing.

But content quality was higher (and volume lower) at NYT than post. I've been a reader since well before the NYT imprint, and I don't think the quality went down significantly at NYT.


All I read was the headline and I feel like we need to go over some things you may have missed in school...


Yes, last time I checked life has a 100% death rate...


Current numbers indicate humans have a 94% death rate, but we're pretty confident that the last 6% will succumb in time.


I am positively sure that is not the case.

There have been "modern" human beings (homo sapiens sapiens) for at least several tens of thousands of years. Even accounting for the fact that human population today is much bigger than it's historical median level, you would expect that if 94% death rate was anywhere near accurate (and assuming a power law distribution of age), you would have no trouble to find a few people that are 200-300 years old. I don't feel the need to calculate the actual numbers, but I suspect that it would be plausible to believe there was at least one "millenial" (someone older than 1,000 years) alive today.

Instead, what embryology tells us is that the power of life and the power of death are not opposite to each other, but deeply entangled. Your body is what it is in equal parts by the fact that many of your cells reproduced themselves and specialized, but many more committed suicide so that the whole organ could live and thrive. And if you stop cellular death, you don't get immortality, you get cancer.


Just to clarify, keslag was joking.


Well it was a joke but also a numerically accurate joke - estimates are that roughly 6% of all humans ever born are alive today. See for example http://www.huffingtonpost.com/2012/11/07/world-population-ex... or http://www.prb.org/Publications/Articles/2002/HowManyPeopleH...


Also contains an important lesson about how statistics mislead: 6% doesn't imply an even distribution within the sample!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: