Incurred sample reanalysis: enhancing the Bland-Altman approach with tolerance intervals.

ABSTRACT

This article describes the need for incurred sample reanalysis and suggests that the combination of a Bland-Altman plot and tolerance intervals can provide a visual evaluation of method performance. It also shows how the proposed combination is a tool that may be of value in determining minimum sample size. An example dataset is worked through in its entirety so that a reader unfamiliar with the topic can gain sufficient information to analyze their own data. Related topics include the generation of 66.7% tolerance factors, comparison of mean-normalized and log differences and the use of probability plots to evaluate error distributions. The suggestions presented in this article are meant to be a continuation of the ongoing incurred sample reanalysis discussion and, as such, further comment is invited.