2016 Postmortem
Related: About this forumTrump is promoting one specific poll to claim he's ahead, but
"Each day's poll respondents are a subset of the UAS election panel, roughly 3000 U.S. citizens who were randomly recruited from among all households in the United States."
http://www.latimes.com/politics/la-na-pol-usc-daybreak-poll-methodology-20160714-snap-story.html
"Every day, we invite one-seventh of the members of the UAS election panel to answer three predictive questions"
http://96.127.53.23/election/
Does that mean that this poll will be using a sample of 450 people from the same group of 3000 people for the entire election rather than doing a random sample of all possible voters every day? Is that an accepted way of polling? If the initial sample of 3000 people is biased in any way, wouldn't that bias stay with the poll for the entire election cycle?
I do not dismiss negative poll results, but the methodology here seems open to a biased sample.
safeinOhio
(32,688 posts)he'll be using internet polls.
Bernardo de La Paz
(49,002 posts)Loki Liesmith
(4,602 posts)They don't sample new people. They re-contact people from the same survey panel initially sampled. It's not a poll really. The only true random sampling was done cat the beginning.
Bernardo de La Paz
(49,002 posts)I think we agree.
Loki Liesmith
(4,602 posts)My problem with this method is bias: it will likely usually have a +R bias
Loki Liesmith
(4,602 posts)Each day's sample draws from the same initial population on respondents. The panel initially favored Trump so it's unlikely to ever not favor trump
Democat
(11,617 posts)They can never break out of whatever the initial sample was.