TJeanloz
new member
Reged: 02/05/04
Posts: 20
|
|
I've been looking at the results of an online poll (of a political nature, but it really isn't important whether it's Bush vs. Kerry or Armstrong vs. Ullrich), and my question for the statistically inclined is why the results can vary so dramatically.
Statistically, we typically expect a "large" sample to have the same median and mean as all other "large" samples (assuming a normal distribution). The thinking is that if you poll 1,000 people, the result should not be significantly different if you poll 2,000 people.
But in observing online polls of all ilks (even those where one cannot vote twice), there can be large swings in results as the sample grows. My question is: is a large swing theoretically possible in a large, random sample, as it grows larger? I think the only way these swings can be explained is by biased voting (somebody goes to bikefanclub.com and convinces all of his friends to vote for Lance, biasing the sample). Is it possible for such a swing to occur in an unbiased sample?
|
OldEd
contributor
Reged: 02/05/04
Posts: 117
Loc: The Great South
|
|
poll, is that in a real poll the sample is answering blind.
In an online poll, the current results are usually known. This may motivate respondents to vote on the prevailing side to inflate the margin, or conversely to help a "losing" side catch up. Plus there's elements of herd mentality, stampeding the results a certain way on a question that the respondent has mixed feelings about: "I'm not sure, but most people feel this way, so I'll go along." Wild swings can result from any of these factors.
Online polls are psychologically poisoned by their openness.
Edited by OldEd (06/16/04 04:38 PM)
|
Continental
new member
Reged: 02/10/04
Posts: 5
|
|
Another major factor is that the demographics of poll respondents change during the course of the day. An extreme example would be Sunday morning at 10:00 am versus Sunday afternoon at 1:00 pm. Other examples would be east coast versus west coast time zones, people with access to computers during work hours versus those who only have computer access at home.
|
Nev
captain
   
Reged: 05/03/04
Posts: 376
Loc: Never where I want to be
|
|
ttt
|