News for hackers. Stuff that matters.
Re:FBI Reopens Clinton E-Mail Probe Less Than Two Weeks Befo... by cmdrfred on Friday October 28, 2016 @05:26PM (#12820879)
A single phone poll of "a random national sample of 874 likely voters" (1 in every 350 thousand) does not a fact make in a nation of 300 million people. Sample size is just too small. Such a study would have to be repeated with a larger sample.
Re:FBI Reopens Clinton E-Mail Probe Less Than Two Weeks Befo... by Armisael16 on Friday October 28, 2016 @05:45PM (#12821023)
56% ± 3.3% at 95% confidence isn't good enough for you?
Re:FBI Reopens Clinton E-Mail Probe Less Than Two Weeks Befo... by cmdrfred on Friday October 28, 2016 @05:54PM (#12821085)
They are pulling the confidence level out of their asses. The Clinton campaign has been playing with the polling since the start of this campaign.

https://wikileaks.org/podesta-emails/emailid/26551

If this is real, it can be duplicated.

Re:FBI Reopens Clinton E-Mail Probe Less Than Two Weeks Befo... by swang on Friday October 28, 2016 @09:34PM (#12822136)
you understand what that email was talking about right?

oversampling is done to get a better view of specific groups. when put back in comparison to the population as a whole they are readjusted for their extra sampling.

those polls were also for internal purposes to the clinton campaign, and not polls that are actually released. so why would there be manipulation in an internal poll that would affect decision making in the clinton camp?

there is literally zero going on in that email but trump supporters saw "oversampled" and freaked out because they didn't know what it meant in regards to polling.

Re:FBI Reopens Clinton E-Mail Probe Less Than Two Weeks Befo... by Armisael16 on Friday October 28, 2016 @06:03PM (#12821150)
...that confidence level isn't from the campaign. That's just how confidence is calculated. The z score for 95% confidence is 1.96, so the range is:

1.96sqrt(0.56*(1-0.56)/871) ~ 3.3%

Re:FBI Reopens Clinton E-Mail Probe Less Than Two Weeks Befo... by gragas on Friday October 28, 2016 @06:32PM (#12821335)
We aren't talking about the true population though. You can't use z-scores, not that it would make a huge difference.

FiveThirtyEight is extremely biased however, and greatly oversamples Democrats.

Re:FBI Reopens Clinton E-Mail Probe Less Than Two Weeks Befo... by michaelmrose on Saturday October 29, 2016 @01:39AM (#12822768)
They don't actually even do polls they agrigate data from a variety of sources.