Three option poll / margin of error
If I have a three option poll.... do I calculate the margin of error the same as I would for a two-option poll?
Using n, z & p and then apply the same e to each of the three answers?
Also, I am getting mixed signals from z tables.... is z=3.3 the 0.999 CL?
Re: Three option poll / margin of error
Usually the assumption is that the number of votes in a specific option has a marginal Binomial distribution, and the votes in each options have a jointly multinomial distribution.
So if you are interested in the standard error of the estimator of the proportion of a specific individual option, then the estimate is the same no matter how many options you have in your poll.
However, when you are dealing with some combinations of the proportions, e.g. the difference of the proportions between two options, then you have to take care with the covariances.
Advertise on Talk Stats