posted by steven .
at the census bureau web site you can find the percent of adults in each state who have at least a bachelors degree. it makes no sense to find the mean for these data and use it to get a confidence interval for the mean percent U in all 50 states. Why not?
The fraction of people with a college degree in all states would require a WEIGHTED mean of the individual state averages, with the weighting done by population. The populations of states vary by a large factor. California has 50 times more people than North Dakota, for example.