A direct comparison of two slope-estimation techniques used in method-comparison studies.

Abstract

Least-squares regression estimation of the slope and intercept in method-comparison studies is the most common method of data evaluation despite well known theoretical difficulties with this approach. An alternative estimation technique has been suggested, which avoids some of these theoretical difficulties. I developed an exact mathematical relationship between these two slope estimates in order to directly compare their characteristics. A simple approximate formula for the relative difference between the two slope estimates is also derived. Using these mathematical relationships, I compare the two estimation techniques, using data from published method-comparison studies. In most of the published method comparisons the differences between slope estimates were trivial, but in some the differences were large enough to have warranted use of the alternative slope estimate.