Investigating the agreement between methods of different precision

Abstract

Agreement between measurement methods is commonly investigated by a so-called Bland-Altman plot showing if the difference is independent of the size of the measurement. However, such analysis assumes that both methods have the same precision. If not, the plot may show a false trend caused by the difference in precision and not by an actual dependence on the size of the measurement. We suggest a modification of the Bland-Altman plot such that the differences are plotted against an inverse-variance weighted average of the measurements rather than their mean. This study shows that such modification removes the dependence on difference in precision.