Linear Correlations

Abstract

Central to all statistical evaluations of the reliability of results in quantitative laboratory medicine is the correlation of results of testing for analytes on more than one analyzer. In fact, regular performance of correlation studies is a requirement of many regulatory bodies.

There are several approaches to performing correlation studies; the most common approaches used in pathology are two-tailed T-test or ordinary least square best fit line approach. The latter approach aims to fit a line to experimental observations by minimizing the error of each observation point from the line. The product of this approach is a regression line that explains the correlation between the analyzers using the slope and intercept.

A similar statistical procedure used in laboratory medicine is linearity and calibration. Linearity is the term used for the procedure to establish the limits of sensitivity of a given quantitative assay, and calibration is the comparison of test result values with those of a calibration standard.

In this chapter, we have explained the least square best fit line approach and introduced the concepts of linearity and calibration.