--On Thursday, 31 August, 2006 09:38 +0200 Brian E Carpenter
<brc(_at_)zurich(_dot_)ibm(_dot_)com> wrote:
Full disclosure: My personal opinion, which I *did* give to
Lynn and
Andrew when I became aware of this glitch, is that a reset is
the only
way to be certain that the selection process is unbiased.
Brian,
I don't know about others, but I'd like to hear a little more
about your reasoning (and Andrew's) about this. It seems to me
that drawing a second sample would be unbiased if the decision
to draw it were made before anyone knew the contents of the
first sample. But, as soon as someone looks at the first sample
and then has discretion as to whether to say "never mind" and
draw another one, there is bias in the statistical sense. That
bias may or may not be harmful, or have the appearance of being
harmful, but it definitely removes the rigid randomization of a
method that doesn't allow any latitude or individual choice in
the selection of a candidate pool.
To illustrate this, suppose that one initially drew two
membership pools from the list of volunteers. Now examine the
following cases:
(i) Someone looks at the contents of both pools, decides
which one is preferred, and picks that one.
(ii) Someone decides to look at one of the two pools and
then decide whether to accept it or to select the other
pool.
(iii) The second pool is drawn after some or all the
members of the first pool are withdrawn from the initial
volunteer list, with the mechanism for selecting those
who are withdrawn being exogenous to the process and
presumably deterministic.
There are rather complex, and quite intriguing, models in
statistical decision theory for examining each of these types of
cases. But none of them involve "unbiased" with regard to the
randomness of the selection process.
john
p.s. I deliberately haven't looked at the volunteer lists to
determine who the relevant IAB member was, making the comment
I'm about to make unbiased by that knowledge. But I believe
that an IAB member who is sufficiently unfamiliar with our
procedures to have volunteered to the nomcom should be seriously
considering stepping down (which would not make him or her
nomcom-eligible, of course). I also believe that this
micro-debacle suggests that future revisions of the nomcom
selection document should be explicit about two cases:
(1) Sorting the nomcom volunteer pool into alphabetical order
and then assigning numbers that will, in turn, be used in the
determination of who gets selected is not appropriate. The
sequencing of the volunteer pool should probably use a
randomization process that is demonstrably independent of the
randomization process that selects nomcom members from that list.
(2) Just as the rules that link the date of resignation from a
nomcom-selected position with rules about filling the resigned
position (e.g., with regard to duration of terms) need
clarification in some way that can be reviewed by the community
via Last Call, we probably need absolute clarity about the
relationship between the date of resignation and eligibility to
serve on a Nomcom, initiate recalls, etc.
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf