 |
 |
Geert Aufdemkampe writes: >We are at the moment trying to fit more than two raters >in a Bland and Altman plot. Conventional software such as >Medcalc accepts only two raters. Do you have a solution >for this problem? I'm a bit confused here. Usually, the Bland and Altman charts are used to compare two different rating instruments rather than two different raters. For example, you might compare the temperature measured with a mercury glass thermometer to the temperature measured with one of those strips that you hold to your forehead. Raters are different than ratings, because you usually want to generalize beyond the two or three raters themselves. The raters in your study probably cannot come to my hospital to evaluate my patients. I would have to rely on raters that live and work in Kansas City. It might be better to come up with measures of how consistent two or more raters are, such as the intraclass correlation coefficient. There is a nice book by Shoukri and Edge that gives details about how to measure agreement among multiple raters, but the mathematics are rather difficult. If you wanted to compare the three raters like you would compare three ratings, then a simple approach would be to do a Bland Altman chart for A versus B, then for A versus C and the finally for B versus C. Before you do this, though, think about whether this chart has any practical utility. Shoukri, M. M. and V. L. Edge (1996). Statistical Methods for Health Sciences. Boca Raton, FL, CRC Press. Steve Simon, [log in to unmask], Standard Disclaimer. STATS: STeve's Attempt to Teach Statistics. http://www.cmh.edu/stats %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%