hey - if India expelled all the Muslims, Hindus and Sikhs, that should end the problem of communal violence! Population' d shrink a bit though.

And I know how to solve the Israel/Palestine problem now! Etnically clense the Arabs! Can't believe I used to criticise Likudniks.

Eh; they still can't make a decent ham and cheese sandwich. However, you have to be careful about simply eliminating the Arabs; the Druz are staunch allies of the Israelies. Being a despised cult to the Muslims and all, they figured `why not'; and it would be bad form to wipe them out now.

That article probably should be in a tabloid magazine. They said the polled 700 people over the telephone (not a good way to poll people unless you can't really do it any other way). Those questions could have been something like: "Do you think Muslims organizations should be watched or do you think George Bush is a Liberal?"
Plus on top of that, only racist people who actually support taking away civil liberties from Muslims would actually answer those questions. I personally would just hang up.

I love phone surveys! I'm a 93 year old Eskimo orange farmer and koala breeder! Well, just to provide some more information, I went to Cornell University's web site, and pulled this up

http://www.comm.cornell.edu/msrg/report1a.pdf

This makes things a little clearer, though adds a couple of questions on the quality of the survey. I'm no expert in statistics, but my math isn't quite working out the same as theirs. Ah well.

How else are you gonna poll people nation wide? Hire a hundred people, randomly select cities, pay their airfare, and have them pound pavement? The internet? You can't get a random sample on the internet. Phones aren't the greatest method, but we haven't thought of anything better yet.

I have to agree that the headlines sensationalize the results...

Its certainly not as bad as implied. Here's the key table, pulled from teh report:
http://www.misc-debris.us/sluggypics/al ... ic_id=2521

only the first one is really scary.

Really dot?

I found the first 3 rather terrifying. Not surprising (yay for my bitter cynical side) but terrifying. Number 4 is damn scary too, but at least that one could be tempered with only the ones with suspicious practices 'n stuff.

Kea, you just need to get half a dozen people, have them go to half a dozen malls and ask people to fill out a survey for a few bucks. You've got something that is less likely to be biased and probably more random. If you are going to be polling people nationwide (not just 700) then yes, phones are the way to go.
Plus if you happen to call up Bun-Bun... well you all know what would happen.

Jarne, I think the subset of 'people that don't go to the mall' is a heck of a lot bigger than the subset of 'people that don't have phones'. So even with the added incentive to make people actually respond, you're still getting horribly biased responses; in this case, instead of excluding Luddites, you're excluding the poor, the non-capitalistic, and the crowd-haters. I'm pretty sure that group is larger (and I belong to two out of three, and belonged to the third just a few months ago!)

Ok, so going to the mall and polling people would be a bad idea. But my point was that if they're going to start saying that almost half of all americans want to take away the rights of muslims (which isn't true) then you have to start questioning how they got that information and how accurate it is. Phone polls can be easily biased.

Trust me, I had to take a class on this. 0_o Blegh.

Phone sampling is the best way we currently have of doing surveys. You can get a phone directory and select numbers randomly all over the country. It's not perfect, of course. You point out a lot of flaws, and you're right. People hang up on survey takers. The survey takers try to counter this by phoning back several times, at different times of day, which only works to some extent. You also miss the people who only have cell phones. The survey takers recognise these problems, but they haven't thought of anything better yet.

If you randomly stick people in malls, you miss people who don't go to malls. You also only get to interview people in the few cities you decide to target. There's also the problem that your volunteers might only approach people they feel comfortable talking to (ie. some volunteers may avoid men more than women, or avoid teenagers, or people with piercings, or the badly-dressed etc.)

If you do a survey on the internet, you only get the people who are interested in the website it's on. You also only get the people who are bored enough to do the survey. And you miss all the people who don't have internet (which is more than people who don't have phones).

So, as Kea explained, you can't get a perfect sample. So you record as much as you can about each person you talked to, and you try to figure out how good the sample is. Not enough republicans answered their phones? (This is a problem that US phone polls have; republicans are simply less likely to answer their phones.) Weigh the results; make the republican results you did get more important so that the numbers come out as they would if you had had the same percentage of replublicans responding as the percentage of republicans in your target group. This was one of the big problems with the 2004 exit polls; the pieces of paper had the sponsoring logos on them; ABC, CBS, ect. Many republicans now distrust those media conglomerates and so refused to fill out the surveys more than democrats did. And the early exit polls were not corrected for this. Of course, it takes a while to figure out how many republicans were voting and how many replubicans you surveyed and what the difference is. (Sorry republicans; I know not all sampling error is your fault, these are just two good examples I know, one from statistics class and one from the radio.)