Table II.

Parameter estimates for the mixed multinomial logit model fit to Prostate Cancer SEER Data. Each estimate shows MLE and the empirical mean (in brackets) of the MLEs obtained by parametric bootstrap (simulation from the fitted model). Standard errors (SE) and confidence intervals (CI) shown are based on parametric bootstrap. Upper part of the table has fixed effects, lower part shows random effects. Response categories: 1: Stage=LR, Grade=WM; 2: Stage=LR, Grade=PU; 3: Stage=D, Grade=WM; 4: Stage=D, Grade=PU

	Contrasts
Variables	2 vs 1	3 vs 1	4 vs 1
Intercept	-0.509 (-0.500)	2.301 (2.283)	2.901 (2.890)
SE	0.0125	0.0257	0.0236
CI	(-0.524, -0.475)	(2.233,2.334)	(2.844,2.936)
Age	0.0347 (0.0346)	0.0638 (0.0636)	0.0749 (0.0749)
SE	0.00108	0.00201	0.00171
CI	(0.0325,0.0368)	(0.0597,0.0676)	(0.0716,0.0783)
Year	-0.0144 (-0.0145)	-0.0560 (-0.0558)	-0.0585 (-0.0584)
SE	-0.000264	0.000488	0.000542
CI	(-0.0150, -0.0140)	(-0.0567, -0.0548)	(-0.0595, -0.0573)
DT	-0.0375 (0.0376)	-0.153 (-0.152)	-0.249 (-0.249)
SE	0.0340	0.0100	0.0128
CI	(-0.0442, -0.0309)	(-0.171, -0.132)	(-0.274, -0.224)
ScrDx	-0.108 (-0.107)	-3.394 (-3.428)	4-3.741 (-3.761)
SE	0.0270	0.415	0.466
CI	(-0.160, -0.0539)	(-4.241, -2.614)	(-4.674, -2.848)