[Skip to Navigation]
May 1993

Statistical Analysis for Method Comparison Data

Author Affiliations

Berkeley, Calif

Arch Ophthalmol. 1993;111(5):582-583. doi:10.1001/archopht.1993.01090050016009

To the Editor.  —We read the article by El-Maghraby et al1 in the July 1992 issue of the Archives with specific interest in the generic analysis of data on the reproducibility of measurement techniques. We would like to suggest to the authors, the Archives reviewers, and the ophthalmic community at large a simple analysis technique for method comparison data that avoids the pitfalls of the correlation coefficient method employed by El-Maghraby and coworkers.1As outlined by British statisticians Altman and Bland,2,3 a simple plot of the difference between (in this case) the measurements made by two different technicians and the mean of those two measures would provide a more informative, intuitive, and statistically correct analysis of the reproducibility of the laser flare/cell meter. The mean of the differences represents the bias present between the two observers, the two measurement methods, or the two occasions, and the 95%

Add or change institution