On Sat, Mar 13, 2010 at 1:45 AM, Adam Williamson <email@example.com> wrote:
> On Fri, 2010-03-12 at 21:43 -0700, Stephen John Smoogen wrote:
>> Neither can be done without an outside/neutral polling agency
>> contacting and getting responses from at least 600-3000 random Fedora
>> users. The poll that was given was one that could be easily stuffed
>> and not easily proven that it wasn't. Relying on the forum for data is
>> bad science and makes this whole argument more and more farcical. I am
> Again, when I'm trying to be a scientist, I'll be sure and let you know.
> =) The poll wasn't intended to be a statistically valid indication of
> the entire Fedora user base. For the record, I don't think it's
> sufficiently strong to support the claims Kevin is trying to make it
My sincere apologies Adam, I was not trying to make you into the
villian of this, and probably should have worded things differently.
The poll was done in a sincere effort to get information. And some
sort of poll is needed to start seeing what questions need to be
asked. However an initial poll like this only has a confidence level
of around 50% and no ability to do confidence levels (+/-X amount you
see on various polls in the news)
The results from it have been parlayed as facts; when in the end, they
are just a viewpoint of a select view of people (those who register on
the forum, use the forum regularly,etc ), and because of the nature of
the forum it is hard to determine how much validity there is in the
testing. Even if it was not stuffed, the uncertainty of the tool makes
it so that if say the slower updates had been 70% of the population,
the other side would claim it wasn't a valid test (and vice versa).
> I don't think it was stuffed, though, for a couple of reasons. One, it
> wasn't discussed anywhere outside the forums until over 100 votes were
> in (and the percentages then were about the same as they are now). Two,
> when a poll's being stuffed, you usually see a large amount of votes
> arrive in a lump; I've been watching the vote counts for the poll, and
> that hasn't happened, they've mostly dribbled in a few at a time.
> BTW, it would be very difficult to take up your suggestion, as we don't
> *have* a big list of All Fedora Users for the external polling agency to
No we would have to determine it from FAS... probably using the group
CLA and been a member for over X time. You would then poll through
them sending a directed email to a random set and point them to a
closed poll. After the poll had been validated you would then announce
the results (or announce that it had failed validty tests).
While the FAS+CLA is a self selected group, it is representative of
the developers of the group which would fit into the population that
both groups seem to be thinking they talk for
> generate a random list from. All the usable lists of Fedora-related
> people we have probably suffer from some kind of selection bias. The
> best would be the Smolt data, but of course that's not identifiable in
> any way you could use to contact people.
> Adam Williamson
> Fedora QA Community Monkey
> IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
> devel mailing list
Stephen J Smoogen.
Ah, but a man's reach should exceed his grasp. Or what's a heaven for?
-- Robert Browning
devel mailing list