Margin of Error

A group of individuals are randomly sampled from a given population and are asked their preference in terms of two political candidates. Since a random sample of preferences is subject to some level of uncertainty we want to assess whether any observed differences between these outcomes can be considered statistically significant. The margin of error for each candidate separately (which I’ll describe more fully below) can overlap in which case there is a non-zero probability that they are in a draw and we need to compute this probability. The assumption throughout is that the samples are truly random with no compromising bias. That, of course, is a complicated issue about sampling protocols that I do not address here.