Re: Substantial nomcom procedure updates (Was: Re: Consolidating BCP 10 (Operation of the NomCom))

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

 



At 10:36 AM 9/16/2014, Stephen Kent wrote:
>I like Dave's suggestion i.e., reduce the per-company NOMcom appointment
>limit to 1 from 2.
>
>It's not a perfect solution, but it's clear, simple to implement, and
>the intent is obvious.
>
>Steve


As I said, I did take a look at that approach, and it is *better* than the current process for some value of *better".  But it still has the characteristic that by overloading the volunteer pool, a company can pretty much guarantee its representation on the Nomcom.

The formula is basically  "P = 1-(1-n)^10"  where P is the probability of having a member and "n" is the proportion the company has of the volunteer pool.   E.g. A company with 15% of the volunteer pool has an 80% chance of having a member of the nomcom.

Turning the formula around to how many volunteers you need to get a specific result, you get "n = 1-ROOT (1-P, 10)"

So to have a 95% chance of having a member, you need about a 26% share of the Nomcom volunteer pool.


If you cap the share of the volunteer pool (by doing the two stage selection process I mentioned or something similar), you can set the numbers any way you want.  For example, if you cap the share at 10%, and you're selecting at max one member per company, a company has about a 65% chance of having a member.  If you use the same 10% cap on the pool, but allow a max of two nomcom members, a company has about a 33% chance of no member, 33% chance of 1 and 33% chance of two.

I'm not sure what the right numbers are, but I would like to set things up so that overloading the nomcom pool is no longer a viable strategy (or at least has a lower payoff).

Changing the cap to 1 member will reduce, but not eliminate,  over representation of large companies on the Nomcom.  If the change is made, then year to year  2-4 companies will provide about 30% (guesstimate based on what I remember of past volunteer pools) of the nomcom members; down from what I would guess is currently close to 50%.  The question is whether that number is still too high or not?

Later, Mike

Note:  This is all binomial distribution probability stuff.  In excel, the probability pulling exactly N black balls (nomcom positions) out of Y pulls (10 nomcom slots) given a P percentage of black balls (nomcom volunteers) in the pool is =BINOMDIST (N, Y, P, FALSE).  When you're doing Y pulls, and you cap successes at N, then the result is the sum of N to Y of that function (e.g. the probability of pulling at least 2 black balls is the sum of the probabilities of pulling 2, 3, 4...10 black balls).

I used that to play around with various scenarios.

For instance, assume that 4 companies collectively have a 50% share of the volunteer pool and that each individual company is capped at 1 member.  The numbers work out to about an 82% chance the companies will share 4 members,  a 12% chance of 3, 5% of 2 and 1% of 1.  That goes to looking at the nomcom representation on a longer term basis than year to year.  If the cap is 2 (as it has been in years past), that gives you a 27% chance of 0-3members, a 20% chance of 4,  a 25% chance of 5, a 20% chance of 6, 12% chance of 7 and a 5.5% chance of 8.  









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