8. Transformation of Measurements

  1. Bendix Carstensen1,2

Published Online: 24 JUN 2010

DOI: 10.1002/9780470683019.ch8

Comparing Clinical Measurement Methods: A Practical Guide

Comparing Clinical Measurement Methods: A Practical Guide

How to Cite

Carstensen, B. (2010) Transformation of Measurements, in Comparing Clinical Measurement Methods: A Practical Guide, John Wiley & Sons, Ltd, Chichester, UK. doi: 10.1002/9780470683019.ch8

Author Information

  1. 1

    Steno Diabetes Center, Gentofte, Denmark

  2. 2

    Department of Biostatistics, University of Copenhagen, Denmark

Publication History

  1. Published Online: 24 JUN 2010
  2. Published Print: 13 AUG 2010

ISBN Information

Print ISBN: 9780470694237

Online ISBN: 9780470683019

SEARCH

Keywords:

  • Bland-Altman plot;
  • data transformation

Summary

If a simple Bland—Altman plot of differences versus averages between methods shows signs of non-constant variance, one possible way of compensating for this heteroscedasticity is to find a suitable data transformation that stabilizes the variance. When the variance is increasing with the measurement values, the first remedy is to transform the original data by the log. The transformation of measurement values is an external thing relative to the models and procedures discussed; transformation is applied first, the relevant model fitted to the transformed data, and in the reporting the back-transform is applied to the plot. If more than two methods are compared we may in principle select different transformations in order to make relations linear with constant variances.

Controlled Vocabulary Terms

Bland-Altman plot; Data transformation