Statistical methods in medical research. II. Sample sizes required in experiments involving all-or-none responses.