[EL] McDonald study, birthdate distribution in real voter list
Michael McDonald
mmcdon at gmu.edu
Sun Sep 11 17:04:45 PDT 2011
I provided the Birthday Problem pointer merely to explain why you need to
remove the unmatched birthdays/birthdates when constructing the probability
calculation. The principles are exactly the same between the two probability
calculations, assuming a uniform distribution of days, be they in a single
year or across multiple years. The only difference is the number of days one
is calculating the probability of a match over: 365 versus some multiple of
365. In the article, we go into more depth about the variants that we may
want to consider if we relax the uniform date assumption in the double
voting context. In fact, we explore empirical distributions drawn from the
2004 New Jersey voter registration file and the 2004 Current Population
Survey. We also calculate the expected number of matches, instead of just
the probability of a single match. You're right that this is more involved
than a Wikipedia entry. If you're interested in the details, just follow the
provided link on SSRN to the entire article.
============
Dr. Michael P. McDonald
Associate Professor, George Mason University
Non-Resident Senior Fellow, Brookings Institution
Mailing address:
(o) 703-993-4191 George Mason University
(f) 703-993-1399 Dept. of Public and International Affairs
mmcdon at gmu.edu 4400 University Drive - 3F4
http://elections.gmu.edu Fairfax, VA 22030-4444
-----Original Message-----
From: Paul Lehto [mailto:lehto.paul at gmail.com]
Sent: Sunday, September 11, 2011 7:46 PM
To: mmcdon at gmu.edu
Cc: law-election at department-lists.uci.edu
Subject: Re: [EL] McDonald study, birthdate distribution in real voter list
On 9/11/11, Michael McDonald <mmcdon at gmu.edu> wrote:
> Second, here is the Wikipedia entry for the Birthday Problem. Although we
> are interested in the Birth Date Problem for the Double Voting
application,
> the principles why the non-matched dates are removed are discussed here.
> This is a very common statistics puzzle posed to students, and we cite in
> our article many published articles by statisticians who have explored
> variants to the Birthday Problem, some of which people are raising.
Really,
> this is a very well-understood puzzle that may be difficult to wrap your
> head around the first time you are presented it, which is why it is often
> presented in introductory statistics classes. And, perhaps, why it lends
> itself to double voting allegations.
> http://en.wikipedia.org/wiki/Birthday_problem
I'm no big fan of the double voting "problem" but at the same time I
don't believe the self-deprecating assertion above that Michael
McDonald's post-doctoral work here is really as straightforward as the
birthday problem in Intro Stats (which I did take and passed back in
the mid-80s). Having the same name and the same date of birth is so
much a "variant" of the birthday problem it's just as fair to call it
a different problem. For one thing, the whole field of statistics is
inapplicable (except as a guesstimating process) because neither dates
of birth nor names are truly randomly assigned, thus defeating the
foundational assumptions of normal statistics. Consequently, reading
wikipedia (if seriously meant as a source here) or refreshing one's
self on introductory statistics can not and will not lead to all the
answers.
However, a proper simulation under controlled conditions might give
more nearly correct answers, if the methodology is proper. So might
an actual full scale investigation tracking voter registrations back
to unique individuals. The very fact that the results here are
advertised as "surprising" or counter-intuitive also suggests that the
results might be due to incorrect methodology. Thus, I think we are
back to the original questions, knowing that "Introductory Statistics"
is clearly not the level at which Professor McDonald's work is being
done. It appears the questions are being deflected rather than
answered. Perhaps the true answers are sufficiently complex that they
don't lend themselves to listserv discussion. I'd be more inclined to
accept that over a reference to wikipedia and introductory statistics,
even if it is Professor McDonald himself who appears to be asserting
that his work is readily understandable by reference to such basic
things.
Paul Lehto, J.D.
> ============
> Dr. Michael P. McDonald
> Associate Professor, George Mason University
> Non-Resident Senior Fellow, Brookings Institution
>
> Mailing address:
> (o) 703-993-4191 George Mason University
> (f) 703-993-1399 Dept. of Public and International Affairs
> mmcdon at gmu.edu 4400 University Drive - 3F4
> http://elections.gmu.edu Fairfax, VA 22030-4444
>
>
> -----Original Message-----
> From: Bev Harris [mailto:bev at blackboxvoting.org]
> Sent: Sunday, September 11, 2011 6:16 PM
> To: Paul Lehto
> Cc: mmcdon at gmu.edu; law-election at department-lists.uci.edu
> Subject: Re: [EL] McDonald study, birthdate distribution in real voter
list
>
> Actually, you wouldn't remove the non-matching birth date, just as you
don't
> remove the possibility of landing on heads from the probability of coin
toss
> #3
> just because coin #2 landed on heads.
>
>> My question on the quoted portion from Michael McDonald's email above
>> goes to why would one "have to remove that [non-matching] birth date
>> from the [remainder of the] matching search?"
>
>
> Bev Harris
> Founder - Black Box Voting
> http://www.blackboxvoting.org
>
> * * * * *
>
> Government is the servant of the people, and not the master of them. The
> people, in delegating authority, do not give their public servants the
right
> to decide what is good for the people to know and what is not good for
them
> to
> know. We insist on remaining informed so that we may retain control over
the
> instruments of government we have created.
>
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
> _______________________________________________
> Law-election mailing list
> Law-election at department-lists.uci.edu
> http://department-lists.uci.edu/mailman/listinfo/law-election
>
--
Paul R Lehto, J.D.
P.O. Box 1
Ishpeming, MI 49849
lehto.paul at gmail.com
906-204-4026 (cell)
View list directory