On Tue, May 4, 2010 at 11:14 AM, Kevin Kofler <kevin.kofler(a)chello.at> wrote:
Stephen John Smoogen wrote:
> Any one of those can invalidate the mathematical tests you say to run
> as they require random pools, controls on populations polled, and
> non-leading questions. People keep telling you this and you seem to
> keep ignoring it.
I know the poll is far from perfect. But it is the best we have and it is a
better basis for decision than somebody's guess which has no roots in
reality at all.
And unlike your motorcycle example, there is no provable bias into either
direction. All you can say is that the results MAY be false due to imperfect
methodology, you can't prove they are.
I am sorry, but I was confused. You said you would compute p-value
tests, but those are only really valuable if you have over a 90%
confidence level in the data. Using a non-random sampling polling
method which the forum comes in.. I was advised that the best
confidence level one could have was 60% with it probably being more
like 50%. At this point your +/- % is about 50% also.
At this point, what I can say about the survey is that somewhere
between 1% to 100% of people want 'adventurous updates'. And between
1% and 100% do not. Yes the number 78% looks really strong but the
confidence in it is so low that it could just as easily have been 22%
.. and just as been accurate.
--
Stephen J Smoogen.
“The core skill of innovators is error recovery, not failure avoidance.”
Randy Nelson, President of Pixar University.
"We have a strategic plan. It's called doing things.""
— Herb Kelleher, founder Southwest Airlines