A multivariate hierarchical Bayesian approach to measuring agreement in repeated measurement method comparison studies

Table 2 Estimates of bias, B _{(y, z)}, and the associated 95% credible regions (95% CR) together with estimates of the 95% limits of agreement for pair-wise comparisons of systolic blood pressure measured by the two observers and the automated machine (m = J, R and S) from two multivariate hierarchical Bayesian models.

	J vs. R	J vs. S	R vs. S
	Mean (95% CR)	Mean (95% CR)	Mean (95% CR)
Bias, B _{(y, z)}
HB ₁	0.08 (-0.21, 0.37)	-15.6 (-19.7, -11.6)	-15.7 (-19.8, -11.7)
HB ₂	0.09 (-0.20, 0.37)	-15.4 (-19.5, -11.2)	-15.5 (-19.5, -11.3)
95% limits of agreement
HB ₁	(-4.36, 4.56)	(-56.2, 25.1)	(-56.0, 24.5)
HB ₂	(-4.39, 4.57)	(-55.9, 25.0)	(-55.7, 24.6)
Within subject covariance of measurements between methods $τ_{(y, z)}^{2}$ *
HB ₁	35.5 (28.5, 44.1)	16.1 (7.7, 26.6)	17.4 (8.9, 27.1)

*Note: larger within subject covariance values are desirable.
HB ₁ which assumes that the underlying values remain static and uses the raw data, x _mir, and HB ₂ which assumes that the underlying values can continually change and uses paired difference data, d _{(y, z)ir}.

ISSN: 1471-2288