3 ms·
I don't know how robust their sampling method is, but how many samples do you need in a survey to do that level of extrapolation? I was under the impression tha
by hobs 11y ago
I don't know how robust their sampling method is, but how many samples do you need in a survey to do that level of extrapolation?
I was under the impression that the political surveys extrapolated with less than this.
- cdumler 11y agoThe answer is: it depends. I remember the lesson my Stats teacher: How many Martians do you have to measure to determine the average height of a Martian? What if I told you that all Martians are exactly 6 feet tall? The correct way to sample to is to determine the variance in the population. If there is no variance, sample size needs to be only one. If not, there are statistical charts that tell you what your sample size needs to be to have some specific confidence that some percentage of the time your sample will be representative. If your population has a Normal distribution, you can use a Z chart to calculate sample size. Believe it or not, for normal distributions and command variances, a population below 200 is all that is necessary.
- kencausey 11y agoThere's a difference between random dynamic sampling and fixed sampling. Not that I think sampling is ever done very well in surveying, not that it's the most serious problem.