Re: Update

[Mark Allman <mallman@grc.nasa.gov>]

> I am not sure these numbers being close tells us much about who is
> on the list.  I.e.  just because the numbers are close, it does
> not mean that the same people are on the lists.  I would be
> curious to know how many names were common to both lists.

A little exercise...

  * we have a history of the last 5 meetings; from this we can
    determine someone's meeting attendance pattern

  * so, there are 2^5 or 32 different patterns

  * 13 of the patterns yield no eligibility under either the "3 of
    5" or "2 of last 3" criteria 

  * 13 of the patterns yield eligibility for nomcom under either "3
    of 5" or "2 of last 3" 

  * 3 of the patterns yield eligibility if using "2 of 3" but no
    eligibility if using "3 of 5"

  * 3 of the patterns yield eligibility if using "3 of 5" but no
    eligibility if using "2 of 3"

  * that is, 16 patterns yield eligibility for "3 of 5" and 16
    patterns yield eligibility for "2 of 3"; and there is
    significant overlap in those two sets (nearly 70%)

Now assume that attendees fall into the 19 patterns that yield
eligibility under one of the constraints fairly evenly.  We would
expect to see the pool size be roughly equivalent under either "2 of
3" or "3 of 5".  And, that is what Harald showed us.  So, at least
in terms of back-of-the-envelope thinking the above may be roughly
accurate enough as a model to think about.

So, by choosing "3 of 5" what we are really saying is that we like:

  + to ensure that someone has participated in at least a year's
    worth of meetings, as opposed to 2 meetings

  + we think the people fitting in the 3 patterns in the "3 of 5"
    set and not the "2 of 3" are more useful in nomcom than the set
    of people fitting in one of the 3 patterns that pass the "2 of
    3" rule and not the "3 of 5" rule

My own opinion is that "3 of 5" is better than "2 of 3" for the two
above reasons (especially the min).  But, the difference is in 15%
of the pool -- and, I don't think that matters terribly much in the
end.  Even if you go to something like "4 of 8" you still see
significant overlap with the "2 of last 3" group (again roughly
70%).

So, maybe we are sweating the small stuff here.  I have said before
that I think we should do something fairly different.  But, that has
zero consensus.  That's fine.  So, maybe it is time to Just Pick
Some "M of N" - with the knowledge that what we're really setting is
the lower bound (i.e., how many meetings until eligible) rather than
really gauging one's attendance record over time.

allman


--
Mark Allman -- BBN/NASA GRC -- http://roland.grc.nasa.gov/~mallman/