BF69
join:2004-07-28
Camden, TN | Pure BS
A what is "broadband" do they use the governemnt standard of 200 Kbps? That's glorifed dial-up in my book.
Also you can't tell anything from 1600 people. Maybe 16,000 then I'd take a poll more seriously. |
|
WayneHolmes
join:2006-12-23
Mountlake Terrace, WA
| Assuming the same sampling methodology, increasing the sample size will reduce the margin of error.
We can argue about how the sample is determined, but for that conversation to be significant would require that we both have Doctorates in Statistics or Mathematics. I don't, so will defer to anyone with the appropriate mathematical background to talk about sample determination. (I haven't used those neurons since college. )
In case people are interested, here is a link that offers some fairly understandable background on issues around margins of error and sampling techniques.
»www.wku.edu/~david.neal/statisti···ml#Table
Wayne |
|
Ahrenl
join:2004-10-26
North Andover, MA
 ·Verizon FIOS
| The reduction in margin of error going from a sample population of 1,600 to 16,000 is statistically insignificant. Let's pretend that it's a linear relationship (which it's certainly not). That would mean that the error would move from 2.5% to .25%, and the mean would move the difference of 2.25%... It would be relatively the same result.. |
|
BF69
join:2004-07-28
Camden, TN
| said by Ahrenl  :The reduction in margin of error going from a sample population of 1,600 to 16,000 is statistically insignificant. Let's pretend that it's a linear relationship (which it's certainly not). That would mean that the error would move from 2.5% to .25%, and the mean would move the difference of 2.25%... It would be relatively the same result.. I think 2.25% is alot. if your pay got reuced 2.25% I'm sure you'd notice.
Anywas 1600 is to enough. One could poll 1599 from say NY City adn 1 from Backwoods Ark. And that's a valid survey of the nation? Of course not. 16,000 would make a more even distibution more likely. There are over 3500 counties in the US so at best you aren't even surveying half of them. And only polling ONE person from Chicago, NY or LA is not valid either.
You can not make any valid statement on 300,000,000 based on 1600. Going by that they most likely polled a whopping 30 people in my state. And as I said of all 30 are in say Nashville of course you're going to get a high% saying they have broadband. If they polled 30 people from Holaday the answer will be 0%. Neither paints an accurate picture. |
|
Nuts
join:2006-04-27
Forest, OH
| Now, BF, don't let facts get in the way of a good story. 
BTW, in 2000 just over 59 mil people were in rural America, that was 20.99%. I'm sure that percentage has gone down since then. |
|
BF69
join:2004-07-28
Camden, TN
| said by Nuts :Now, BF, don't let facts get in the way of a good story. BTW, in 2000 just over 59 mil people were in rural America, that was 20.99%. I'm sure that percentage has gone down since then. Once again what is "rual" what is "broadband"? What criteria and methods that where used in the polling. Surely they just didn't ask "Do you have access to broadband internet"? Sampling errors rate doesn't mean shit if the sampling is wrong from the start.
As far as the 59 mil? perhaps. I do know that 225 million Americans live in the top 200 metro areas in the US. Whether you want to call #201 "rural" when it contians over 200,000 people is up for debate. |
|
Ahrenl
join:2004-10-26
North Andover, MA
·Verizon FIOS
| reply to BF69
Well it's not. Would you care if the story read 51% or 55% instead of 53%? Because that's what we're talking about here. So If you like, just tell yourself, between 51-55%.
Weather or not the 1600 was properly selected goes back to the character of the pollee. Which is why I stated that you needed to start there. If they're truly an independent organization, with no agenda (pretty unlikely, because who does a poll like this without an agenda?) then they can easily select a representative 1600 out of 300m. This is pretty basic statistical theory. Of course you can game it, but if you're actually trying to get an accurate reading, then it can be done pretty easily. |
|
