Table 2.

Results for five ordinal GLMM models fitted to Beam et al’s mammography study [2]. Each ordinal GLMM varies with regards to inclusion of rater and subject characteristics, including patient’s age (young = 0, old = 1) and rater inexperience (experienced = 0, inexperienced = 1).

		Model (i)	Model (ii)	Model (iii)	Model (iv)	Model (v)
Parameter	Symbol	Estimate (se)	Estimate (se)	Estimate (se)	Estimate (se)	Estimate (se)
Ordinal GLMM:
Thresholds: ( ${\alpha }_0=-\infty ,{\alpha }_5=+\infty$)
Between categories 1 and 2	${\alpha }_1$	−0.897 (0.135)	−0.902 (0.135)	−0.827 (0.138)	1.095 (0.666)	1.133 (0.670)
Between categories 2 and 3	${\alpha }_2$	−0.197 (0.135)	−0.201 (0.135)	−0.127 (0.138)	1.795 (0.666)	1.833 (0.670)
Between categories 3 and 4	${\alpha }_3$	0.761 (0.135)	0.757 (0.135)	0.831 (0.138)	2.753 (0.667)	2.791 (0.670)
Between categories 4 and 5	${\alpha }_4$	2.539 (0.137)	2.535 (0.137)	2.610 (0.140)	4.531 (0.667)	4.569 (0.671)
Fixed Coefficients:
Subject’s age	${\beta }_1$				0.034 (0.012)	0.034 (0.012)
Rater Experience (inexp =1)	${\beta }_2$					0.055 (0.094)
Random Effects Variance Components:
Subject intercept	${\sigma }_{u0}^2$	2.442 (0.288)	2.442 (0.104)	4.615 (0.543)	0.687 (0.081)	0.687 (0.079)
Subject’s age slope	${\sigma }_{u1}^2$			0.00145 (0.0002)	0.00014 (0.00002)	0.00014 (0.00002)
Rater intercept	${\sigma }_{v0}^2$	0.158 (0.023)	0.135 (0.019)	0.158 (0.023)	0.158 (0.023)	0.135 (0.019)
Rater’s inexperience slope (inexp=1)	${\sigma }_{v1}^2$		0.218 (0.030)			0.159 (0.022)
Fleiss’ kappa for multiple raters	κF	0.297 (0.001)