Trump is promoting one specific poll to claim he's ahead, but

[Democat]

"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]

1. When that one fails

he'll be using internet polls.

[Bernardo de La Paz]

2. It's a rolling 7 day average poll. Drop oldest day data, sample 450 people today, and average. . .nt

[Loki Liesmith]

3. That's actually incorrect

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]

4. We are both correct. I worded badly. I should have written "questioned" rather than sampled.

I think we agree.

[Loki Liesmith]

6. Gotcha

My problem with this method is bias: it will likely usually have a +R bias

[Loki Liesmith]

5. It's a recontact survey

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]

7. That's what it seems like

They can never break out of whatever the initial sample was.

[pnwmom]

8. Also, the poll was conducted among citizens, not registered voters. nt