For multiple sclerosis patients we wish to estimate the mean age at which the disease was first diagnosed. We want a 90% confidence interval that is 8 years wide. If the population variance is estimated to be 85 from previous research, how large a sample should be taken?