hello all,
so, nick and i thoroughly confused ourselves on the last section of
question one. the way that we did the other ones is by taking sqrt(1/n)
as an estimator of standard deviation.
doesn't it follow that to get a wider confidence interval, you have to
survey FEWER people? and yet the wording of the question implies that Eli
is spending a lot of money surveying huge numbers of people, but that the
sample still has to be random. thus he's not doing cluster or stratified
sampling or anything like that, but he's still getting either a very
skewed result, or alternatively a very wide confidence interval.
can anyone help?
cheers,
zoe and nick