Extended biasplot command to assess bias, precision, and agreement in method comparison studies

Recently, a new statistical methodology to assess the bias and precision of a new measurement method, which circumvents the deficiencies of the Bland and Altman (1986, Lancet 327: 307-310) limits of agreement method, was developed by Taffe (2018, Statistical Methods in Medical Research 27: 1650-1660). Later, the methodology was extended to assess the agreement. In addition, to allow for inferences, simultaneous confidence bands around the bias, precision, and agreement lines were developed (Taffe, 2020, Statistical Methods in Medical Research 29: 778-796). The goal of this article is to introduce the extended biasplot command, which implements these latest developments, and to illustrate its use by applying it to simulated data included with the command. Note that the Taffe method assumes that there are several measurements by one of the two measurement methods and possibly as few as one measurement by the other for each individual. The repeated measurements need not come from the reference standard but from any of the two measurement methods. This is a great advantage because it may sometimes be more feasible to gather repeated measurements either with the reference standard or the new measurement method.

Automated summary

A new statistical methodology has been developed to assess the bias and precision of a new measurement method, which overcomes the limitations of the Bland and Altman limits of agreement method. This methodology has also been extended to assess the agreement between measurements. Simultaneous confidence bands have been developed for the bias, precision, and agreement lines to allow for inferences. The extended biasplot command is introduced in this article to illustrate the use of these latest developments with simulated data.