Tweet
What does the saying "the survey has a margin of error of plus or minus 4 percent" mean?
-
-
what is a margin of error
I think that much of that is based on two things. 1) The chance that someone will answer a question based on peer opinion rather than their actual opinion. 2) Depends on what the surveyor wants it to reflect. (During the last presidential election, CNN kept changing it's MOE to be the lead of Mr. Bush. If he had a 4 percent lead, the MOE was +/- 4%, and one time he had a 8 % lead and they changed the headings on their website to reflect +/- 8% MOE.) -bret -
what is a margin of error
What does the saying "the survey has a margin of error of plus or minus 4 percent" mean? Surveys are based on sampling, and margin of error refers to the margin of sampling error. It's a mathematical calculation based on the size of the sample as compared to the entire population of whatever it is you are sampling. If you sampled EVERYONE, there would be 0 margin of error, in other words. The larger the sample in proportion to the TOTAL, the smaller the margin of error will be. If a survey says that 75% of people believe xyz, and that survey has a margin of error of 4%, it means that the ACTUAL number will be 71% - 79% if you took the time to ask everyone. Get it? You literally add and subtract the margin of error to get the range of what the actual result should be if everyone was counted. This also explains the term "statistical dead heat". Suppose this survey was to determine the winner of the next presidential election. If the survey said Dubya has 42% and Gore has 47% with a 4% margin of error, that's a statistical dead heat. Why? Because Gore's actual number could be anywhere between 43% - 51%, while George Jr.'s could be as high as 46%. Either could win, in other words, so it's a statistical dead heat. HTH.Comment
-
what is a margin of error
I don't mean to nitpick, but margin of error is based purely on sample size, not on the proportion of sample size to population. For example, in opinion surveys, you might hear that the result is accurate to within 5%, 19 times out of 20. For such a result, you need a sample size of slightly less than 1000, regardless of the total population size. Notice the phrase "19 times out of 20", which many reporters leave out when reporting on opinion surveys. This means that 1 time out of 20, the survey is totally out to lunch, and you never really know at the time that you have that instance! Again, one more thing to remember when reading opinion survey results. (This is largely from the two statistics courses I had to take in university 20 years ago.)Comment
Comment