Re: IAOC requesting input on (potential) meeting cities

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



I meant to include (damn iPad):

Carsten's original email was
Statistically, the chance of any IETFer being killed last week was about 1.5 %.

"Any IETFer being killed" (I read the above as "Any [given] IETFer") is the .00051% chance. "At least one IETFer" is the .5% chance. "Anyone in the IETF" could mean either - precision is important and I still don't actually know which Carsten meant.

Later, Mike

On 4/5/2017 9:49 AM, mike stJohns wrote:
Let's try this one more time


27 deaths in 100 would be 27%
27 deaths in 1000 would be 2.7%
27 deaths in 10000 would be .27%
27 deaths in 100000 would be a rate of .027%
The above are all expected deaths per year for the given population - so divide .027% by 52 to get .00051% per week.  That's roughly the chance per individual to die during the week from murder.   That's *my* risk for a week of the IETF and that's what should matter to each individual.

If you take that and multiply it by the IETF population you get .005 expected deaths within the IETF for the entire week.

Now you're using the "at least one" criteria, which works out to be 1-(1-.0000051)^1000 = .005 or .5% of at least one person dying in the week so your numbers are somewhat reasonable, but don't actually mean anything.  E.g this is basically the birthday paradox which grows the probability quickly with the size of the set.

  I'll give you an example.  The death rate per 1000 in Germany per year for all causes is about 11.  That works out to about a 20% chance of at least one person dying in the IETF during the week from all causes including murder if we all lived in Germany.   The equivalent statistic for the US is 8 per 1000 which works out to 15% for the IETF for the week.  But we don't seem to be dropping at anywhere close to that rate ( e.g. A death at the meting every 5 meetings).

Death probabilities are *highly* dependent on the specific populations you're looking at and actuaries get paid a lot of money to calculate them.  The actual probabilities for the IETF would benefit from affluence, age, medical care, education, location, etc.

Mike



Sent from my iPad

On Apr 5, 2017, at 01:56, Carsten Bormann <cabo@xxxxxxx> wrote:

On Apr 5, 2017, at 01:06, mike stJohns <mstjohns@xxxxxxxxxxx> wrote:

.005 deaths per week per 1000
Right, the 270 micromorts per year (~ 5 micromorts per week) you cited would lead to a ~ 0.5 % chance of anyone of the ~ 1000 IETFers getting killed — a factor three less than the 1.5 % that my numbers result in, but not off by a large factor.
(Still, I wonder where that factor three comes from.)

Luckily, we hit the other 98.5 % this week.

(I’m not that worried by the ~ 15 micromorts of risk I personally was subjected to, given that I spent on the order of 250 microlives for the time that went into this meeting.
But my wife, who had been planning our wonderful vacation in Colombia earlier this year, definitely was.)

Grüße, Carsten





[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]