Posted on 09/11/2005 7:51:28 AM PDT by new yorker 77
Final Topline Results (9/10/05)
N = 1,009 national adults, 18 and over
Margin of error: plus or minus 4
Interviewing dates: September 8-9, 2005
SAMPLE SIZE/MARGIN OF ERROR FOR REGISTERED VOTERS/
SUBGROUPS:
901 Registered voters (plus or minus 4)
SAMPLE SIZE/MARGIN OF ERROR FOR KEY SUBGROUPS:
300 Republicans (plus or minus 7)
334 Democrats (plus or minus 6)
331 Independents (plus or minus 6)
------------------------------------
793 Whites (plus or minus 4)
195 Non-whites (plus or minus 8)
You're also not making much of a point, unless you can identify a statement I made that you disagree with. Because it's not very clear what you're disagreeing with me over.
My information comes from my research class in college. The instructor, who had been through the process of defending a dissertation (meaning he was Dr. Smith) told us as a class that once the sample size goes below 10% of the population for which you are applying the results, then the accuracy starts to deteriorate. If you think about it, it makes sense. Can 1000 adults really tell you what Americans are thinking as a whole when there are nearly 300 million Americans? I just don't think they can. However, 29 million randomly selected adults would definitely be a more accurate representation of Americans.
OK..........then I'll simplify for you.
You asked me to speak for the "original poster," and I told you that I could only speak for myself.
It's not that complicated.
Now, back to the point. You believe a Newsweek poll. The rest of us are smarter than that.
Good bye now.........
But in reality, it doesn't. If you do the math, you'll find that if you want to take a snapshot of the views of 290,000,000 people (which represents the entire population of the United States, not just adults let alone likely voters) a random sample of only 385 people will give you a picture of their views that, 95% of the time lie within 5% of reality. If you want to reduce the margin of error to only 2%, you need a sample of 2401. If you want 99% certainty and a 2% margin, you still only need 4147 respondents. Obviously the more certain you want to be and the least margin you want to have, the more respondents you need, but you can have a high degree of confidence with a relatively low number of respondents.
There ARE many issues with polling reliability, which include the fact that pollsters aren't getting sufficiently representative samples because normal people don't answer telephone surveys. Other pollsters poll at the wrong time, or ask the wrong questions, or even screw up their math.
That's a good point I hadn't thought of before. :-)
You're right, it's not that complicated. All I was looking from you is a statement I made on the subject that you disagree with. The most you could come up with was the fact that you didn't want to speak for another poster. Nothing at all that pertains to the subject of discussion. The only conclusion is that you don't disagree with anything I said there. Thank you for confirming that.
When the fog lifts in your brain, get back to me.
(It's not often that a freeper communicates as poorly as you do, but has the nerve to blame his own density on someone else. I'll remember that and be watching for you).
(Just like a lot of the Jr. High boys I once taught. I loved them too........and some of them actually grew up eventually ;).
Against my better judgement, I went back and read every thing you posted on this thread, just to verify what I thought was true. (It was).
You didn't say anything. All you did was argue with the poster of the thread, and ask some really silly questions. (you think only 29% of voters are Republican? LOL!)
There was no substance to what you posted. Just trust in the poll and questioning people who are clearly more educated than you are about the way these polls work.
Of course I said something. I explained why the poster hadn't made his case.
you think only 29% of voters are Republican?
As was pointed out by the original poster, not all registered voters actually vote in any given election. More than 29% of those who voted in 2004 were enrolled as Republicans (that number is 37%) - no one disputes that. The question is whether significanly more than 29% of registered voters are Republicans.
Since the number of registered voters is substantially greater than the number of people who actually vote, and since unenrolled people are expectedly less likely to vote than those who are enrolled in a party, it would follow that the percentage of people who are enrolled in either party is going to be higher among those who vote than among those who are registered to vote. You following?
"The reader" will see the little winky face ;) at the end of my comment and know that I was just teasing you, inquest. Sorry you missed that.
As for your last post, thanks for actually saying something finally.
I've got a life outside FR, so I can't read it carefully right now and respond, but I will later.
Thank you. I would like to point out for the record, though, that what I said at #112 is substantially the same as what I said at #11 and #46. But I will be looking forward to your comments.
I am skeptical to say the least. Sample size is everything. I don't believe the poll can be accurate with numbers that small.
Take elementary college statistics, my numbers are correct. The real art is, of course, finding a representative sample. But if you are able to do that, you'd be surprised at how few respondents you need to get an answer with a high degree of certainty.
Of course, it isn't feasible to poll 290,000,000 Americans, or even 29,000,000. Therefore, I take all poll results with a grain of salt. They may or may not be accurate.
I view this idea of using a small sample size and having it be an accurate representation as a theory. It can never be prov-en, because you do not know what those who were not asked would say.
On the contrary, it can be proven by simple laws of probability. The question comes down to how certain you want to be. If you want to be 95% certain the answer falls in the margin of error (which is the standard), you're still allowing that one in twenty polls falls outside the margin of error. Moreover, large margins of error make polls meaningless, because they indicate trends that may not be happening.
The 10% figure came from a professor in a research class who had gone through the process of defending a dissertation. It's an established benchmark. I don't know where it came from, but it certainly makes sense. Polling less than 1% and saying it represents everyone doesn't make sense. That's just my take.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.