if no preliminary sample is taken to estimate p, how large a sample is necessary to be 87% sure that the point estimate will be within a distance of 0.04 from p?