A researcher desires to estimate the mean serum cholestreol level for college age women. In a past study the standard deviation for this measure was 42 units. If the researcher wants to be off in his estimate by no more than three units (with 99% confidence), how large must his sample be?