Disagreement Plots and the Intraclass Correlation in Agreement Studies
Although disagreement and the intraclass correlation have been covered previously, several variants of both have been proposed. This chapter introduces readers to several variants of the disagreement plot and the classification of the intraclass correlati
- PDF / 236,598 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 40 Downloads / 207 Views
Disagreement Plots and the Intraclass Correlation in Agreement Studies Suhail A.R. Doi
Abstract Although disagreement and the intraclass correlation have been covered previously, several variants of both have been proposed. This chapter introduces readers to several variants of the disagreement plot and the classification of the intraclass correlation coefficient and the concept of repeatability in agreement studies.
Variations of the Limits of Disagreement Plot Several variations have been proposed for the limits of disagreement plot (also called a Bland–Altman plot). In terms of the x-axis, instead of the average we could use a geometric mean or even the values on one of the two methods, if this is a reference or gold standard method (see Krouwer 2008). In terms of the y-axis, the difference (d) can be expressed as percentages of the values on the average of the measurements (i.e. proportional to the magnitude of measurements). This helps when there is an increase in variability of the differences as the magnitude of the measurement increases. Ratios of the measurements can be plotted instead of d (avoiding the need for log transformation). This option is also utilized when there is an increase in variability of the differences as the magnitude of the measurement increases. All these can be done with the help of routine software such as MedCalc, which gives a warning when either the ratio or percentage includes zero values.
S.A.R. Doi (*) Clinical Epidemiology Unit, School of Population Health, University of Queensland, Brisbane, Australia Princess Alexandra Hospital, Brisbane, Australia e-mail: [email protected] S.A.R. Doi and G.M. Williams (eds.), Methods of Clinical Epidemiology, Springer Series on Epidemiology and Public Health, DOI 10.1007/978-3-642-37131-8_3, © Springer-Verlag Berlin Heidelberg 2013
29
30
S.A.R. Doi
Fig. 3.1 Regression of DXA based %body fat versus PLETH based %body fat demonstrating a good linear relationship. However this does not necessarily mean there is good agreement
Fig. 3.2 The classic layout of the Bland–Altman plot comparing %body fat measurements by DXA and PLETH in terms of limits of disagreement. The limits suggest that measuring % body fat by PLETH could result in an absolute difference of 2.9 % to +7.9 % compared with DXA
Example Sardinha (1998) designed a study to compare air displacement plethysmography with dual-energy X-ray absorptiometry (DXA) for estimation of percent body fat (%BF). The differences between DXA-determined and plethysmographydetermined %BF were compared (Figs. 3.1, 3.2, 3.3, 3.4, and 3.5).
The Special Case of Limits of Disagreement for two Successive Measurements: Repeatability Coefficient The repeatability coefficient tells us the maximum difference likely to occur between two successive measurements when we want to know how good a measurement instrument is.
3 Disagreement Plots and the Intraclass Correlation in Agreement Studies
31
Fig. 3.3 The % average layout of the Bland–Altman plot comparing $body fat via DXA with %body fat via PLETH i
Data Loading...
