 |
SwiftVets.com
Service to Country
|
| View previous topic :: View next topic |
| Author |
Message |
ocsparky101
PO1
Joined: 03 Sep 2004
Posts: 479
Location: Allen Park. Michigan
|
 Posted: Sat Oct 23, 2004 1:55 am Post subject: Can the polls be trusted? |
 |
|
| According to Dick Morris, "these pollsters have never had to deal with an election that is projecting an over 60% voter participation. Therefore the poll results are based on a voter turnoput of 53% or less turn out". That means the GW could possibly have a much more substantial lead than what is being shown in the polls. |
|
| Back to top |
|
 |
cipher
Vice Admiral
Joined: 10 Aug 2004
Posts: 902
|
| Posted: Sat Oct 23, 2004 2:02 am Post subject: |
|
|
That's a rather specious argument.
If 53% of a statistical sample of 50% of the population that eats ice cream say they like vanilla, 53% of a sample of 60% is going to still be 53% liking vanilla, even if it is a larger number.
Then again, 75% of people polled say they lie to pollsters....
_________________
USMC 69-72, 7th Comm, 3rd MarDiv, FMFPAC
US Army 75-79, 97th Sig, SHAPE, NATO
Arkansas National Guard 79
Defense contractor for US Navy, SSPO, SP-20, SP-24, OP-12 84-92 |
|
| Back to top |
|
|
The Balloon Artist
PO3
Joined: 25 Aug 2004
Posts: 262
Location: Texas
|
| Posted: Sat Oct 23, 2004 2:10 am Post subject: |
|
|
| cipher wrote: |
Then again, 75% of people polled say they lie to pollsters.... |
Who took that poll and how do you know they aren't lying now?
_________________
What about John Kerry's four months in Vietnam qualify him to be president?
Al Gore was there for five. |
|
| Back to top |
|
|
Anker-Klanker
Admiral
Joined: 04 Sep 2004
Posts: 1033
Location: Richardson, TX
|
| Posted: Sat Oct 23, 2004 2:13 am Post subject: |
|
|
Factor this into your thinking. Pollsters are like everyone else; they have a service to sell. In this crazy election season we're seeing more pollsters than I think we've ever seen before, so there's a lot of competition. The polls we see are bought by the news media, and both of them have a vested interest in keeping the "news" content high, i.e., suspence and drama are "good," same-oh-same-oh is bad.
There's probably no monetary advantage, under the circumstances, for any pollster to say that one candidate or the other is leading by a sizable margin consistently.
You'll have to decide for yourself how much you believe them, but I'll argue there's a conflict of interest problem here. |
|
| Back to top |
|
|
RogerRabbit
Master Chief Petty Officer
Joined: 05 Sep 2004
Posts: 748
Location: Oregon
|
| Posted: Sat Oct 23, 2004 2:17 am Post subject: |
|
|
http://jaycost.blogspot.com/2004/10/just-how-bad-are-polls-lets-measure-it.html
If you believe this - then they proably cannot
| Quote: | Just How Bad Are Polls? Let's Measure It!
Yesterday I said that now that we have so many national polls, we can finally make a few claims about the real state of the national race. Some people have responded that we can make practical guesses with fewer polls, and I want to quantify exactly how much of a "guess" we are making each time around.
To do this, I want to look at state polls. Suppose that for each state we have two state polls with historically reliable methodology and with a 4% margin of error. Also suppose that the polls do not round their polls to 0 decimal places.
Note that there are at least 12 states (NM, CO, OR, NV, WI, MN, IA, MI, PA, OH, FL, NH) where one candidate has a lead somewhere between 0 and 5%
Here are some questions to answer:
Polls In A Tied Race
Q: If the candidates are actually tied in a state, what is the chance that both polls will show one candidate with the lead?
A: 24.6%
Q: If there are twelve states that are actually tied, what is the chance that both polls for at least one state will show one candidate with the lead?
A: 96.6%
Polls In a One Point Race
Q: If one candidate has a one point lead in a state, what is the chance that at least one poll will incorrectly call the leader?
A: 40.52%
Q: If one candidate has a one point lead in a state, what is the chance that both polls will incorrectly call the leader?
A: 16.4%
Q: If there are twelve states where a candidate has a one point lead, what is the chance that the polls will at least split on the leader in at least one state?
A: >99.999%
Q: If there are twelve states where a candidate has a one point lead, what is the chance that both polls for one state will inaccurately call the leader in at least one state?
A: 88.3%
Polls In a Two Point Race
Q: If one candidate actually has a two point lead in a state, what is the chance that at least one poll will incorrectly call the leader?
A: 31.21%
Q: If one candidate actually has a two point lead in a state, what is the chance that both polls will incorrectly call the leader?
A: 9.7%
Q: If there are twelve states where a candidate has a two point lead, what is the chance that the polls will split on the leader in at least one state?
A: >99.99%
Q: If there are twelve states where a candidate has a two point lead, what is the chance that both polls for one state will inaccurately call the leader in at least one state?
A: 70.8%
Polls In a Three Point Race
Q: If one candidate actually has a three point lead in a state, what is the chance that at least one poll will incorrectly call the leader?
A: 23.27%
Q: If one candidate actually has a three point lead in a state, what is the chance that both polls will incorrectly call the leader?
A: 5.4%
Q: If there are twelve states where a candidate has a three point lead, what is the chance that the polls will split on the leader in at least one state?
A: 99.8%
Q: If there are twelve states where a candidate has a three point lead, what is the chance that both polls for one state will inaccurately call the leader in at least one state?
A: 48.7%
Polls In a Four Point Race
Q: If one candidate has a four point lead in a state, what is the chance that at least one poll will incorrectly all the leader?
A: 16.35%
Q: If one candidate actually has a four point lead in a state, what is the chance that both polls will incorrectly call the leader?
A: 2.8%
Q: If there are twelve states where a candidate has a three point lead, what is the chance that the polls will split on the leader in at least one state?
A: 98.7%
Q: If there are twelve states where a candidate has a four point lead, what is the chance that both polls for one state will inaccurately call the leader in at least one state?
A: 27.8%
Polls In a Five Point Race
Q: If one candidate has a five point lead in a state, what is the chance that at least one poll will incorrectly call the leader?
A: 11.1%
Q: If one candidate has a five point lead in a state, what is the chance that both polls will incorrectly call the leader?
A: 1.2%
Q: If there are twelve states where a candidate has a five point lead, what is the chance that the polls will split on the leader in at least one state?
A: 94.1%
Q: If there are twelve states where a candidate has a five point lead, what is the chance that both polls for one state will inaccurately call the leader in at least one state?
A: 13.8%
Bottom Line
This is why I do not trust polls, ladies and gentlemen. Bookies might use these numbers, but they are wholly and completely outside the realm of legitimate/objective analysis.
The chances that any given poll is calling the state for the wrong guy ranges between 11% for a 5 point lead to 40.5% for a 1 point lead. Thus, the inaccuracy of polls range from the chances of you flipping heads three times to a little less than the chances of you flipping heads just once.
This makes the marginal value of a poll next to nothing. For the races most people are interested in -- WI, IA, NM, OH -- there is a 20-40% chance that each and every poll will be flat-out wrong not just in giving you the accurate percentages in each state, but also in telling you who is winning.
That means that the more states you examine, you are bound, BOUND, to run into a wrong poll. It is inevitable, as Agent Smith might say.
Furthermore, the more states you examine, the larger your chances of finding a unanimously incorrect call, from 13.8% for a five point lead to 88.3% for a one point lead.
Statisticians know this, ladies and gentlemen. They know that media polls (i.e. polls that are done with extreme infrequency) are crap for predicting elections. They know that to get a good sense of a race you have to do TONS of samples (probably have about 15-20 current samples for every state in question). They release these to make news, not to report the news.
|
_________________
"Si vis pacem, para bellum" |
|
| Back to top |
|
|
cipher
Vice Admiral
Joined: 10 Aug 2004
Posts: 902
|
| Posted: Sat Oct 23, 2004 2:20 am Post subject: |
|
|
| Quote: | | Who took that poll and how do you know they aren't lying now? |
It was the same pollster that said "4 out of 5 dentists surveyed recommend Dentyne for their patients who chew gum"
As to the other, they were pre-qualified by a screen filter asking if they had ever lied to a pollster.
_________________
USMC 69-72, 7th Comm, 3rd MarDiv, FMFPAC
US Army 75-79, 97th Sig, SHAPE, NATO
Arkansas National Guard 79
Defense contractor for US Navy, SSPO, SP-20, SP-24, OP-12 84-92 |
|
| Back to top |
|
|
msindependent
Vice Admiral
Joined: 26 Aug 2004
Posts: 891
Location: Colorado
|
| Posted: Sat Oct 23, 2004 2:44 am Post subject: |
|
|
| Hell, no. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
