In defence of the correlation coefficient

The relationship between calibrated range, residual standard deviation and correlation coefficient r is discussed. It is shown that in typical chemical calibration applications with appropriately distributed calibration points (particularly, with range comparable to mean and with approximately even or with ‘successive dilution’ spacing), the linear correlation coefficient has valid application as a routine criterion for acceptable fit if used with due care.

Stephen L. R. EllisonEmail: Phone: +44-20-8943-7325Fax: +44-20-8943-2767