100 DUers (2 per state) X 15 phone calls each = A poll we can trust!

[ParanoidPat]

So what's stopping us? :shrug: :evilgrin:

Many of us already take the Zogby Survey poll so what's to stop us from 'borrowing' the format and questions to conduct our own telephone survey to check the results published by Zogby?

It could be fun and edumacationable. :) What should we call it?

[quaoar]

1. Statistically irrelevant?

[ParanoidPat]

3. How so?

Do you know how many people are included in the average poll? :evilgrin:

[quaoar]

6. It's not numbers

It's weighting.

[ParanoidPat]

7. Care to explain?

Are you telling me we can't conduct a poll just as accurately as anyone else? Why would that be? A random telephone survey is just that, random.
You can argue all you want about how to weight the numbers after we collect the raw data. :)

[quaoar]

18. No, I'm not saying that

But if the survey isn't demographically valid, it won't mean much.

Two per state, for instance, is not a good idea. That gives California the same weight as Wyoming.

Your survey idea would be great if it can be made statistically valid.

I'm not trying to rain on your parade.

[ParanoidPat]

21. What is a 'statistically valid' approach......

......to conducting a national poll? We need a framework to start with and I am NOT a polling expert! :)

I just figure we have people in every state and by sharing the load we could start doing our own polling. I think we should call it the TIA (Truth in America!) poll. ;-)

[quaoar]

22. Any poll

needs to reflect the electorate -- racially, geographically, income-wise, etc.

You don't want a poll that has 65 percent of the respondents female if women only make up 51 percent of the nation.

[ParanoidPat]

23. I understand that but.....

.....how the hell do they weight a poll of 1000 respondents for all of those factors and still get a MOE of +/- 3 or 4%? What do you do, call 20,000 people until you get just the right mix?

[Joanne98]

2. I'm in Ohio

I don't have free long distance though. But count me in anyway.

[faithnotgreed]

4. i thought the same thing this afternoon. i thought if we got 500

of us which we should be able to do out of this group... then each could make 5 calls or less.. for a large enough sample

and i thought we should weight it a bit more towards reg republicans.

[ParanoidPat]

5. I'm not sure about how or why you would 'weight' the poll.......

.......but I would make the calls at random and take data on registration, party affiliation and likelihood that they will vote.

Then you can weigh the results on the basis of who responded. :)

[faithnotgreed]

14. thats a good thought. i initially figured mine would serve 2 purposes:

one, it would give an idea of how the republicans are tilting, and also we would stack the deck slightly against us, giving i believe a more accurate picture.

we are pretty certain that more republicans are voting kerry than democrats voting bush

but again, this was an initial opinion, and totally unscientific!

[Guaranteed]

8. We'd need more than that per state.

That we give us the 30 needed for each state if we randomize it, but we'd need more like 100 for it to be really reliable.

This is a GREAT idea, though.

[Renew Deal]

10. It could be fun.

[Guaranteed]

24. kick...anyone? anyone? nt

[Renew Deal]

9. Talk to Truth is all

There are also some other bright math people here.

[ParanoidPat]

12. Trust me, I posted this with him in mind!

:evilgrin:

[LeftCoast]

11. By treating each state equally

wouldn't you be strongly favoring small population states far out of their proportion in terms of impact on the race?

on edit: I see others have brought up the 'weighting issue'.

[ParanoidPat]

13. Well if enough of us stopped wasting time fretting over other polls.....

......and started conducting our own, we could conceivably do state polls individually and roll them into a national poll.

[LeftCoast]

15. I'm not criticising your effort...

...Just trying to add what I sort of remember from Stat 101, 8 years ago. :)

I see no reason why DU could not produce a volunteer survey that is scientifically valid. You're basicaly talking about open-sourcing a poll. I think it's a great idea.

[ParanoidPat]

16. Well it beats reading all the DOOM & GLOOM poll threads!

:evilgrin: :toast:

[LeftCoast]

19. Cheers to that!

:toast:

[Momma Mia]

17. ParanoidPat ...

To paraphrase your words somewhat, if I may ...

"Stop wasting time fretting over ANY of these polls!"

The only ones that really count are the exit ones taken on November 2nd.!!!

Don'tcha think, huh?

Just let all the hullabaloo from the RNC calm down for awhile and keep your chin up girl!

We're gonna make it! :thumbsup:

[ParanoidPat]

20. What exit polls are you referring to?......

......IIRC, VNS was disbanded after they screwed the 2000 and 2002 exit polls and were sold to the Saudis by the media consortium that owned them. :(

Is there someone else doing exit polls now or might I be trying to set us up to do it? :)

[dreadneck]

25. only valid poll

The only scientifically valid poll is the election. The rest is a weird sort of mathematical theatre masquerading as social science.

[IdaBriggs]

26. I made my living analyzing numbers for a while.

Still do it somewhat. I'll help with that portion. :)

[David Zephyr]

27. Do It By Zip Codes or Area Codes. It Would Be A Better Start.

I think you might be on to something, Paranoid Pat.

I have worked with professional pollsters and the concerns given above about statistics are legit, but should not present an obstacle to your idea.

For our purposes here, it would be helpful to either use telephone area codes or zip codes.

I'd suggest a larger sampling than just 5 calls though.

[ParanoidPat]

28. I never suggested just 5 calls......

.....what I posted was 2 people per state (each making as many calls as it takes to get) 15 respondents each for a total of 30 per state times 50 states or 1500 respondents. :)

If we can start out small with just a few simple questions we can tweak it as needed and eventually build it up to wherever it takes us.