Friday, February 18, 2005

Who are these guys?

Instapundit has a post today about a poll done by Rasmussen Reports which apparently shows a decrease in approval of the UN. I'm not familiar with Rasmussen Reports, but I'm not familiar with a lot of things.

Thirty-seven-percent (37%) of Americans have a favorable opinion of the United Nations. That's down from 44% in a November survey.

I just want to point out this little blurb from lower down on their website.

Data from the current survey is not directly comparable to last November's results. The current survey sampled American Adults. Last November's survey interviewed Likely Voters. It is likely that the difference in samples may shift the results two to three points in one direction or the other.

Nothing personal guys, but then why did you just compare it at the top of your little article thingy? Some people might not understand why you can't compare these results. So a breif explaination of why not. (See this if the word Sample is scary sounding to you)

When they surveyed Likely Voters in November they randomly selected individuals who were listed as likely to vote. (I think these people are usually registered to vote an possibly voted in the last election but the definition varies). This time they took a sample of individuals who are simply American Adults. Well Likely Voters are a specific subset of American Adults and they may have differences from American Adults that would cause them to respond differently to questions in this survey. "But wait, aren't samples full of lots of different kinds of people? Why would it matter if you added a few more different people to the mix?" It matters because it is possible that Likely Voters are systematically different from the 40% (Guess?) of American Adults who are not Likely Voters. Maybe Likely Voters watch the news more, or maybe they are all more educated or less angry, than the rest of American Adults, I don't know, but it might have caused them to have answers to the survey that were very different.

So if the difference is 7% right now, and it might shift 3 percent that could put us at a difference of 4% or 10% and since the margin of error is 3% the difference could actually be 1% or 13%. (I should point out that both of these aren't as likely as the actual percentages they gave, I don't have enough info to give you an exact probability for both of those possible results but it is pretty low) Aren't statistics fun?

Another side note, if you look at the questions, depending on the order the questions were asked in, it might result in different answers.

If someone asked me "Do you think that Bribes from the UN Oil for Food Program caused nations to protest the US led invasion of Iraq?" I'd probably say no. But if someone asked me if I thought the bribes existed and THEN asked me the above question, I might lead myself to a logical conclusion. That could work either way depending on the knowledge that the individual had at the time. These are tricky questions and so the results aren't completely reliable.

No comments: