I saw a headline today on Reuters: "Obama has slight lead over Clinton in Nevada." The poll in the article showed Obama polling at 32%, Clinton at 30% with a margin of error of 4.5%.
I thought for a minute about if I presented data at conference and the title of my talk was "Compound O is more efficient than compound C." If I presented data saying that O is 32% efficient and C is 30% efficient and my margin of error was 4.5% then the title of my presentation is a complete lie and my credibility as a scientist, an impartial observer of phenomena, would be shot. The best I could say is that O and C are equivalent, nothing further.
Maybe they do new math in the media? I'm all about education though so for the people at Reuters I suggest you re-learn math. Best to start at the beginning so start with this lesson, "My Hero, Zero."
When was the last time the news media covered elections as anything but horse races or sweepstakes? Journalists aren't interested in reporting facts because they don't sell papers/TV progams, and they probably don't understand statistics, anyway.
In this case, you're right to imply that, to oversimplify, the poll results mean the race is too close to call, but to know for sure, the reader would need a lot more information, like the size of the sample--usually, the more people sampled, the lower the margin of error. 4.5% M of E sounds like a small sample. Also, how was the poll conducted? If it was a telephone poll, it'll ignore those who have only cellphones, many of whom tend to be younger and more tech savvy than those with, say, only landlines.
But one can't assume that the 4.5% could go to one or the other, though it might. More likely, it would be divided between the two. Also, margin of error is meant to correct for sampling and non-sampling errors, but they cannot account for people lying to pollsters, interviewers leading a respondent to an answer, etc.
Which comes back to survey size and sampling method being crucial bits of information for evaluating any poll. The more people surveyed using diverse methods, the more likely the poll is to be somewhat correct. But in this election, polls are meaning very little, so it seems to me the whole system is fundamentally flawed. I bet the news media won't report that, either.
Posted by: Arthur (AmeriNZ) | January 15, 2008 at 01:11 AM
Facts? We don't need no stinking facts!
Posted by: homer | January 15, 2008 at 10:41 AM
Facts? We don't need no stinking facts!
Posted by: homer | January 15, 2008 at 10:42 AM
Glad she got Nevada now? I sure am! :D
Posted by: Sorata | January 20, 2008 at 08:40 AM