PDA

View Full Version : Is it possible to make a scientific online poll?




JasonM
09-25-2007, 02:00 PM
Why couldn't an online poll take the same properties of a real poll?

All you'd have to do is have a popup that first asks if you would be interested in a poll. This popup, however, would not appear every time you go onto the page. Rather, there is only a 10% chance that the system will approve your particular IP address.

If the participant says "yes", then there would then be a number of pre screening questions such as whether you are registered to vote, what state you are from, gender, race, etc.

It would also make sure that you are from a US IP as well.

In this way, not just any old person can participate, but only randomly selected people.

Surely there's a way to set this up at a high traffic site somewhere.

MsDoodahs
09-25-2007, 02:03 PM
Still would not be random.

Ridiculous
09-25-2007, 02:07 PM
People who answer phone polls are people with nothing better to do. I mean, we have already established that a large number of people on here don't have landlines.

But of the people that do, most people have caller ID now and aren't going to answer a call from some anonymous number. And then you have to get past all the people who are just going to hang up on you right off the bat.

So you are left with old people with nothing better to do.

JasonM
09-27-2007, 04:00 AM
How would it not be random? I mean, heck, if the fact that "people are choosing to respond" makes it not scientific, what makes the gallop poll scientific? I mean, people are still choosing to respond to a poll. They have to agree with it.

In the example I gave, people would not know about the poll unless they visited the site, were lucky enough to get the popup, AND answered the preliminary questions correctly.

Keith
09-27-2007, 04:31 AM
Fewer and fewer polls are scientific these days. Remember what makes a poll scientific. It's the scientific method.

1. Start with a hypothesis, the media selected front runner should win.
2. Test they hypothesis with some sort of a poll.
3. Ron Paul actually won.
4. Throw out the results because the wrong person won so it must not be scientific.

That is the scientific method right?

Its not?

OptionsTrader
09-27-2007, 04:38 AM
The results of a properly designed and properly implemented poll can be extrapolated to represent a much larger population.

BUT, requirements including that the sample set is a random sample of the entire set means that an online poll is never going to be random, so that simple necessary condition is not going to be met.

The answer to the question of this thread is:

No.

maxmerkel
09-27-2007, 04:47 AM
i'd say it's possible - some ideas :
- poll has to be on different HIGH traffic web-sites that work together (msn, yahoo, google, maybe even msnbc, times, wash. post etc)
- poll is prominently placed at the main page
- poll is only visible 3 random times during the day for at most 3 minutes on each webpage.


that would be VERY scientific representing the average internet user - and i would really be interested how ron paul would fare in such a poll ! i think now it would be 5-10%, in a month 8-15% ad i january at least 30% :D

OptionsTrader
09-27-2007, 04:55 AM
Read this list and notice how many of these are violated in an uncontrolled online poll such as the one described:
http://www.publicagenda.org/polling/polling_20q.cfm

JasonM
09-30-2007, 04:03 AM
i'd say it's possible - some ideas :
- poll has to be on different HIGH traffic web-sites that work together (msn, yahoo, google, maybe even msnbc, times, wash. post etc)
- poll is prominently placed at the main page
- poll is only visible 3 random times during the day for at most 3 minutes on each webpage.


that would be VERY scientific representing the average internet user - and i would really be interested how ron paul would fare in such a poll ! i think now it would be 5-10%, in a month 8-15% ad i january at least 30% :D

Bingo.

That's what I was trying to get at.

The randomness would come from the fact that not everyone would see the online poll. In fact, not everyone would even know it's there unless they're the one within the range of lucky IP addresses which are acceptable (which, effectively, means that you also couldn't use a proxy to vote).

In other words, this isn't your normal garden variety online poll.

american.swan
09-30-2007, 04:19 AM
People who answer phone polls are people with nothing better to do. I mean, we have already established that a large number of people on here don't have landlines.

But of the people that do, most people have caller ID now and aren't going to answer a call from some anonymous number. And then you have to get past all the people who are just going to hang up on you right off the bat.

So you are left with old people with nothing better to do.

The Internet is no better because it's full of another segment of the population. That is the problem. How do you get a proper cross-section of America into one poll unless you do some really hard searching.

First you have to determine what percentage of people even care about politics at all.
Then you have to fine enough potential voters, which is what Ron Paul is after from each demographic.

The On-line demographic is not the same as the voting populous and the on-line demographic that cares about politics isn't correct either. The best POLL is the exit polls.

american.swan
09-30-2007, 04:26 AM
Read this list and notice how many of these are violated in an uncontrolled online poll such as the one described:
http://www.publicagenda.org/polling/polling_20q.cfm

very good read. Seconded.

dude58677
09-30-2007, 05:29 AM
Fewer and fewer polls are scientific these days. Remember what makes a poll scientific. It's the scientific method.

1. Start with a hypothesis, the media selected front runner should win.
2. Test they hypothesis with some sort of a poll.
3. Ron Paul actually won.
4. Throw out the results because the wrong person won so it must not be scientific.

That is the scientific method right?

Its not?

It's pseudoscience. Any test where the results proves the contrary is ignored or explained away.