PDA

View Full Version : Need Help from a Statistician

patriotcalendar
01-13-2008, 12:53 AM

1) A random group of 28 people is asked one question (one time). The question is multiple choice, with six possible responses.

What is the statistical probability that only ONE answer (of the six options) will surface among the entire group? (Assume the possibility that not ALL persons answered.)

2) In a random group of 28 people, six different fruits are discussed.

What is the statistical probability that 18 or more people would raise their hands when asked "how many of you dislike apples?"

Thanks up front for any help anyone can offer!

sw33tbabiblu3
01-13-2008, 03:26 AM
For the first question I assume you are talking about the Luntz focus group? He didn't ask, "Which candidate do you support?" but rather he asked, "How many of you support Fred Thompson?" Since the question is about a specific person, and not all people raised their hands, than it is quite possible that more than 1 answer surfaced, and some in the group simply did not raise their hand.

Anyway, if you want an answer as to the probability of everyone liking one answer of six, the probability of success is derived by taking the probability and multiplying it to the "nth" power, or number of trails.

(1/6) to the 28th power, or approximately 1.628x10^-22 (AKA: Most nearly impossible)

However, we must take into account that what you are referring to is a public opinion poll and not a random variable.
For example, if we asked the question:

"What would you rather do?"
1. Get burned alive
2. Stick a fork in your eye
3. Jump off a building
4. Drown
6. Eat pie

If this showed to be a random data set with a normal distribution, we would assume that there would be a very small chance that the same outcome would occur 28 times in a row. However, looking at the answers, we see that it is very possible that each person could chose the 6th choice, "Eat pie"

Therefore, I have great reserves about using a completely random statistical approach to this question.

For question number 2, the same p-value applies, less than .001 percent , but I still have the same serious caution mentioned above about making this type of analysis.

coyote_sprit
01-13-2008, 03:34 AM
And don't forget these people are watching enough Fox news to want to attend a debate hosted by Fox.

UK4Paul
01-13-2008, 04:01 AM
Frank Luntz is the master at manipulating opinion, and he wrote a book about it.

MoneyWhereMyMouthIs2
01-13-2008, 04:57 AM
Therefore, I have great reserves about using a completely random statistical approach to this question.

Great reserves? lol... I'd say completely inappropriate and useless. :) You made a great explanation.

It's fairly obvious that they were told what to say, manipulated by Luntz, or manipulated socially by Luntz and other group members. I don't believe for a minute that we would have seen the same distribution if all of those people were asked seperately. They aren't finding opinions here... they're creating them.

nosebruise
01-13-2008, 05:47 AM

1) A random group of 28 people is asked one question (one time). The question is multiple choice, with six possible responses.

What is the statistical probability that only ONE answer (of the six options) will surface among the entire group? (Assume the possibility that not ALL persons answered.)

2) In a random group of 28 people, six different fruits are discussed.

What is the statistical probability that 18 or more people would raise their hands when asked "how many of you dislike apples?"

Thanks up front for any help anyone can offer!

hahah it doesnt take a statician to tell you thats highly dependant on other factors.

for example... what was the nature of the question? was it opinion? was the answer common knowledge?

for example if you asked "how many apples are in this box?" and the group could directly observe 3 apples, chances are everyone will answer whatever choice "3" was.

you need more information to derive any sort of logical conclusion to those questions.
you can't just pull meaningful stats out of thin air. the thing that makes statistics meaningful or predictable is the information surrounding them.