An alternative statistical test on the differences between two means in real biochemical analysis

A modified t-test which allows statistical evaluation for biochemical analy sis is proposed. This new test may deal with real cases in which the total error of practical results often has contributions from both systematic and random bias. About 95% of test results fall within the acceptance interval of +/- 15% around true values to meet the requirement of the Washington Co nference Report. The statistical algorithm of power function and required s ample size for this new test were also developed. The performance and limit ation of the method were studied. (C) 1999 Elsevier Science B.V. All rights reserved.