Table 5.

Analysis of variance for 72 renal transplant patients measured twice on ECW (by KBr dilution and MF-BIA) supplemented with estimation of variance componentsa

Sources of VariationSSdfMSEMS = σ2ICC/Proportion
ICC = method agreement = proportion of patient VC to the total amount of patient VC plus the patient-by-method VC,
or: ICC = σ2P/(σ2P + σ2P × M)
For ECW: σ2P = (MSP - MSP × M)/2
σ2M = (MSM - MSP × M)/72
σ2P × M = MSP × M
SDD = smallest detectable difference; with two repeated measurements and a probability of 95%
SDD = √MSP × M × √2 × 1.96
For ECW: SDD = 3.55, so with a probability of 95% score differences lie between -6.8 and 0.3.
“Limits of agreement” are almost equivalent to SDD limits. Factor “patients” is considered to be a “random” factor, the factor “method” is fixed.
aSS, sums of squares; MS, mean sum of squares; df, degrees of freedom; EMS, σ2 = expected mean sum of squares = (estimated) variance component (VC); ICC, intraclass correlation coefficient.
Patients (P)1396.837119.679.020.846
Method (M)379.761379.76(5.25)
P × M (error)116.48711.641.640.154