All else being equal, if you cut the sample size in half, how does this affect that margin of error when using the sample to make a statistcal inference about the mean of the normally distributed population from which it was drawn