Analysis of method-comparison data.

Method-comparison studies, in which new 'candidate' techniques are compared with an established 'gold standard', feature commonly in optometric research. Investigators often analyse their findings using least squares regression, Pearson's product-moment correlation coefficient and the paired t-test. We describe more appropriate methods which include: the 'limits of agreement', the intraclass correlation coefficient, and Deming's method for estimating a straight line fit. Comparisons between methods are made with the aid of a computer simulation. Computer program listings are provided to assist in implementing the analyses.