ClintCooper2003
(629 posts)
Send PM |
Profile |
Ignore
|
Thu Jan-06-05 01:02 AM
Original message |
|
I was just wondering something. On these exit polls, under each question, there's a letter "n" and then something like "= 2520." Does that mean only 2,520 people answered that particular question? TIA? Help!
|
Carolab
(1000+ posts)
Send PM |
Profile |
Ignore
|
Thu Jan-06-05 01:23 AM
Response to Original message |
thanatonautos
(282 posts)
Send PM |
Profile |
Ignore
|
Thu Jan-06-05 01:35 AM
Response to Reply #1 |
3. Yep, that's how it looks to me. |
|
For example, take the file:
US2004G_3970_PRES04_WE_V_Data.pdf
At the top of the table, in the cell above the questions, it says Number of interviews -2,640.
Then, right underneath the first question is
Are you: (n=2,633) Male/Female.
And other questions are all the same, with n<2640.
It sure seems that n is the total number of responses to the question.
|
ClintCooper2003
(629 posts)
Send PM |
Profile |
Ignore
|
Thu Jan-06-05 01:38 AM
Response to Reply #3 |
4. Thanks. Well, the numbers regarding the question "Did you vote... |
|
in the 2000 Presidential election?" are consistent. As it is broken down, Kerry picks up 50.9% to Bush's 47.9% within this question.
It just kind of concerned me because I thought that if only a small percentage went on to answer some of the other questions, it might result in less accuracy for those.
|
TruthIsAll
(1000+ posts)
Send PM |
Profile |
Ignore
|
Thu Jan-06-05 01:44 AM
Response to Reply #4 |
5. Clint, keep this in mind: If n =2500, MOE = 2%; for n=10,000, MOE = 1% |
|
So those responses are very accurate.
N MOE 1000 3.16% 1500 2.58% 2000 2.24% 2500 2.00% 3000 1.83% 4000 1.58% 5000 1.41% 10000 1.00% 11027 0.95% 13047 0.88%
|
thanatonautos
(282 posts)
Send PM |
Profile |
Ignore
|
Thu Jan-06-05 03:44 AM
Response to Reply #4 |
6. Right ... it would affect the accuracy but TIA's table shows you |
|
Edited on Thu Jan-06-05 03:47 AM by thanatonautos
as long as you've got more than 2500 it's still reasonably accurate. The random error is varying inversely as the square root of the number of responders, so the number of responders would have to be cut by a factor of four in order to double the percentage size of the error.
(Edited to fix silly numerical mistake.)
|
Oversea Visitor
(1000+ posts)
Send PM |
Profile |
Ignore
|
Thu Jan-06-05 01:28 AM
Response to Original message |
DU
AdBot (1000+ posts) |
Wed May 08th 2024, 10:34 PM
Response to Original message |