Stable Release Updates types proposal (was Re: Fedora Board Meeting Recap 2010-03-11)

[Stephen John Smoogen]

On Sun, Mar 14, 2010 at 8:14 PM, Rex Dieter <rdieter at math.unl.edu> wrote:
> Peter Hutterer wrote:
>
>> On Sat, Mar 13, 2010 at 09:14:48PM -0700, Orion Poplawski wrote:
>>>
>>> On Sat, March 13, 2010 4:58 pm, Peter Hutterer wrote:
>>> > Isn't there a mere RISK to lose 70-80% of our users if we do _not_
>>> > implement
>>> > the changes as well? Especially given the chance that the poll did not
>>> > represent a significant user sample?
>>>
>>> How many users do we need?
>>
>> sorry, I'm not sure I understand the question. Which user number are you
>> referring to?
>
> I'd venture he meant in response to your "the poll did not represent a
> significant user sample" comment.  So, how many users are needed to make it
> representative?

Well you would want to poll a group that you knew had been active for
a while, and had some certainty that they were unique individuals. I
know several people who register 3-4 accounts on forums so that they
can vote more often or play games. I don't know how many people do
that but the risk usually makes these sorts of polls have lower
confidence levels to around 50%. [EG a statician looking at the data
would say that it is no more valid than flipping a coin.] This is NOT
to say that the poll end data is in the end wrong just that trying to
be more confident in the data (whether it was that Fast, Slow, or
Cowboy Bob) can not be determined. So if the data had shown that
people wanted conservative, I would hopefully be saying the same
things if people said it proved their point.

A more valid poll might be possible by the following method:

If we assumed that the people who had been registered in FAS for over
6 months and had signed the CLA met the first two definitions, you
would need to randomly select about 3000 of them and have at least 600
answer the poll to have (i think) a 90% confidence level in the poll.
I think the questions need to be simple yes/no ones to qualify for the
'easiest' tests, multiple choice results require something like
multiple asking worded slightly different or some such thing. Again
this is from a class I took 20 years ago so a real mathematician,
psychologist, etc would know better.

I also do not want to imply or say that this is wanted/possible etc.
There would be a whole range of issues that would have to get approval
or checked to see if assumptions were valid.

-- 
Stephen J Smoogen.

Ah, but a man's reach should exceed his grasp. Or what's a heaven for?
-- Robert Browning