Danger Will Robinson!
First, please let me be clear -- I am not in any way
disparaging Mr. Root's campaign. He is one of the
candidates who is performing well on the issues most
important to Outright, and is, in my personal opinion,
one of the top-tier candidates.
However, I have to point out that the data from these
"surveys" are completely GIGO (garbage in, garbage
out). I have pointed this out to Stephen Gordon
repeatedly, and am very disappointed to note that he
has chosen to completely ignore these points. Thus, I
really have no choice but to point it out in public.
Firstly, Stephen's poll violates the first law of
statistically significant and population-reflective
polling, since it is self-selecting. Only polls that
survey a statistically-significant random sample can
accurately reflect the perspectives of an entire
population. Stephen's poll is not a random sample
poll and is thus completely invalid.
A simple way to test whether a poll is valid or not is
to ask "if a particular candidate decided to spam the
poll with his or her supporters, would that poll shift
suddenly and dramatically?" The answer in this case
is "yes."
Let's suppose, for instance, that I entered the race
and had 1,000 of my supporters join the "poll" and
support me for president. Stephen's "poll" would show
that "2/3 of Libertarians support Brian Miller for
president" -- an obviously absurd proposition because
over 2/3 of Libertarians have never *heard* of Brian
Miller.
In short, this is a classic GIGO problem.
Now, for the second major problem -- the statistical
validity. In other words, what is the confidence
interval? What is the standard deviation for each
question? What is the margin of error? Without these
measurements, there's no way to tell if the reported
data even accurately reflects the opinions of the
small sample that Stephen has on his list. For
example, Stephen reported that a majority of
Libertarian party members on his list support the
implementation of a change to the LP's rules to allow
Ron Paul to run for the LP nomination in 2008 -- 50.5%
if I recall correctly. However, if the confidence
interval is 10%, it could mean that only 40.5% of the
people Stephen has polled support that position -- or
that a near super-majority do.
I'm very concerned by the lack of disclosure here. On
Stephen's web site for the poll, there's a disclaimer
stating, in essence, "this is experimental and please
accept our apologies if you're harmed by the
conclusions drawn if they happen to be inaccurate."
However, in the news releases that he's been
releasing, he's not inserted this disclaimer, nor
noted that the poll is not a random sample (and thus
not methodologically valid as a barometer for
measuring the population it claims to represent).
Mr. Root may well be the front-runner in this race,
and if so, I congratulate him. However, this poll is
not going to tell us anything one way or the other.
And while we certainly are starved for data in the LP
in general, bad data like this poll is even worse than
having no data at all.
Cheers,
Brian
--- Bruce Cohen <brucedcohen2002@...> wrote: