Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Birthday Paradox Experiment (2018)

pudding.cool

Post reply on HN
Print view View original

71–74 of 74 posts

Re: The Birthday Paradox Experiment (2018)

#71

Post by bambax »

kqr wrote: post #61
bambax wrote: post #17

Earlier quoted context omitted.

Yes, it's also extremely unlikely. The actual distribution in developed countries is not uniform: there is a spike at the end of September (because many more people make babies at or around New Year's Eve) and a considerable drop on Dec. 25th (because people will avoid that date and provoque the birth some days before in case it might happen). Also, on the site there is a huge spike on Nov. 15 which, incidentally, is…

> The actual distribution in developed countries is not uniform This is a nice story I've heard many times, but is it actually true? Like what are your statistical sources here?

US births by day data is available here:

https://github.com/fivethirtyeight/data/tree/master/births

For the period 2000-2014 the number of births by month, divided by the number of days in a month, is:

   1  31   5 072 588     163 631.87   
   2  28   4 725 693     168 774.75   
   3  31   5 172 961     166 869.71   
   4  30   4 960 750     165 358.33   
   5  31   5 195 445     167 595.00   
   6  30   5 163 360     172 112.00   
   7  31   5 450 418     175 819.94   
   8  31   5 540 170     178 715.16   
   9  30   5 399 592     179 986.40   
  10  31   5 302 865     171 060.16   
  11  30   5 008 750     166 958.33   
  12  31   5 194 432     167 562.32   
     365  62 187 024     170 375.41
The difference isn't huge but it's significant. There were on average 180k birth in Sept. vs 165k in April and 170k in December.

Edit: also, here, a graph per day:

https://www.reddit.com/media?url=https%3A%2F%2Fi.redd.it%2Fm...

(The source is mentioned but I did not verify it directly).

Re: The Birthday Paradox Experiment (2018)

#72

Post by kqr »

bambax wrote: post #71
kqr wrote: post #61

Earlier quoted context omitted.

> The actual distribution in developed countries is not uniform This is a nice story I've heard many times, but is it actually true? Like what are your statistical sources here?

US births by day data is available here: https://github.com/fivethirtyeight/data/tree/master/births For the period 2000-2014 the number of births by month, divided by the number of days in a month, is: 1 31 5 072 588 163 631.87 2 28 4 725 693 168 774.75 3 31 5 172 961 166 869.71 4 30 4 960 750 165 358.33 5 31 5 195 445 167 595.00 6 30 5 163 360 172 112.00 7 31 5 450 418 175 819.94 8 31 5 540 170 178 715.16 9 30 5 399…

The scale on that plot is cut off to emphasise small-ish variations. It is practically uniform.

Re: The Birthday Paradox Experiment (2018)

#73

Post by Balgair »

aroberge wrote: post #69
Balgair wrote: post #64

Okay, that was about as clear as mud to me. Maybe I just need more coffee, but nothing about it led me to understand why there is a 50% chance with 23 people. Can someone else explain this? The odds that someone else shares just your birthday out of 23 people sounds crazy still. It should be 182.5 to get to 50%, right?

It is not that someone else shares YOUR birthday: it is that two people (among those 23) will have the same birthday.

oooooh, got it. That wasn't clear.

Re: The Birthday Paradox Experiment (2018)

#74

Post by bambax »

kqr wrote: post #72
bambax wrote: post #71

Earlier quoted context omitted.

US births by day data is available here: https://github.com/fivethirtyeight/data/tree/master/births For the period 2000-2014 the number of births by month, divided by the number of days in a month, is: 1 31 5 072 588 163 631.87 2 28 4 725 693 168 774.75 3 31 5 172 961 166 869.71 4 30 4 960 750 165 358.33 5 31 5 195 445 167 595.00 6 30 5 163 360 172 112.00 7 31 5 450 418 175 819.94 8 31 5 540 170 178 715.16 9 30 5 399…

The scale on that plot is cut off to emphasise small-ish variations. It is practically uniform.

The point of the plot is to show the drop on Dec. 25th; but the numbers are unrelated to the plot and do show actual variations from month to month.
Post reply on HN