Hays Medical Center wants to estimate the percentage of adults who exercise at least 1 hour per week. How large a sample should be selected in order to estimate this proportion with a 99% confidence level and a margin of error of at most 3%? (Assume no preliminary estimate of p-hat is available.)