> Confidence intervals are based on both the sample size and the standard deviation. And since we don't know the standard deviation, we need to estimate it fro

> Confidence intervals are based on both the sample size and the standard deviation. And since we don't know the standard deviation, we need to estimate it from the sample. So we need a lot of samples for that. > Given the 100 mile radius of 3 out of 4, you can have some confidence that this is not random. All the "100 mile radius" tells you is that its concentrated in areas with programming companies. It's not surprising that there are no contributions from Nebraska. > It is quite likely to change with new samples, but it is not meaningless. If the (unknown) probability of a corporate donation coming from the east coast is p, and we have N samples, then the chance that at least N-1 come from the east coast is p^N + choose(N, 1) (1-p) p^N. When N=5, if p = 0.5 then there's an 18.75% chance that we'd see 4 out of 5 donations from the east coast, just by chance. For it to be significant at the 1% level, you need p = 0.28. So even if the west coast were 3 times as likely to donate as the east coast, you would still have 4 / 5 corporate donations from the east coast 1% of the time.