Surgical treatment of coronary disease.

dissemination of Gaussian-based "signifi- cance" tests.If or when the percentile method of expressing interindividual differences has been widely adopted, there will still remain two much more difficult problems: (a) the collection, from healthy and diseased subjects, of data that are reliable and are applicable to the particular patient under investigation; and (b) the question of where to make the cut-off points between "normal" and "abnormal"what risks of error (false-positives and false- negatives) are acceptable.For the balancing of these risks-often a different balance with respect to different disorders, different consequences of error, and even different patients -we must look to clinicians, not to biochemists or other laboratory workers, or to statisticians as such.It would be interesting if the SI units compensated somewhat for their adverse effects by promoting the use of a sensible method of expressing intermeasurement variation and by allowing us to focus attention on how to collect data and apply them to individual patients.