You need to do a little better than what you've been doing. this is almost simulated work. There's no real drift in the stats yet.
The american public has clear and well defined ideas. I am questioning the signal to noise ratio of the poll.
The best technique to employ, chris, is to ask the question three or four different times, innocuously in three or four different places.
Then, instead of trying to get such a big bang for your buck as to have the poll cover so many topics - you just end up answering one, or two questions with the entire poll and you get a very accurate result.
That is, you get them off balance and ask the question four times, then average and weight the results based on their question value and relevance score to the parent question.
So yes, you're effing it up.
But at least, since you're an english major, you can write convincingly about how you're effing it up. Get a math guy there. AN INDEPENDENT. Someone who doesn't give a damn how the poll comes out. thats important.
Or else you'll be delivering a well sharpened axe to grind, fitted just the right size to cut down nothing and cost everything.
PS - Chauvenet's criterion only applies in polls to known value medians. it works best in physical contexts. Still, you might want to look into it, its a generally good way to find out exactly how badly you effed things up.
Oh, and of course - please understand. I'm not saying I don't eff things up. We all do. But right now, I haven't seen anything out of this polling project that looks like anything else other than a fox news version. Those guys are complete idiots. You really have tofind the real beat of the American heart, to be a good pollster. And it is a strong one. Good luck!
