Sequentially calibrating a Bayesian microsimulation model to incorporate new information and assumptions

Table 4 Credible interval overlap measures, area overlap measures, and standardized differences in means based on simulated annual incidence of clinically detected cancers in 1975–1979 (e.g., SEER calibration targets) [27]

	\(p_{orig}(\theta \mid y)\) and \(p_{scratch}(\theta \mid y,z)\)			\(p_{seq}(\theta \mid y,z)\) and \(p_{scratch}(\theta \mid y,z)\)
Clinically detected cancer incidence, 1975–1979	CI overlap	Area overlap	SMD	CI overlap	Area overlap	SMD
Colon, female, 20–49	0.94	0.76	0.31	0.98	0.95	0.02
Colon, female, 50–59	0.95	0.82	0.25	0.98	0.94	0.05
Colon, female, 60–69	0.98	0.91	0.05	0.99	0.95	0.03
Colon, female, 70–84	0.92	0.66	0.48	1.00	0.95	0.02
Colon, female, 85+	0.92	0.70	0.44	0.99	0.92	0.09
Colon, male, 20–49	0.93	0.76	0.33	1.00	0.94	0.03
Colon, male, 50–59	0.96	0.88	0.16	0.98	0.91	0.08
Colon, male, 60–69	0.99	0.92	0.08	0.97	0.90	0.12
Colon, male, 70–84	0.90	0.66	0.47	0.97	0.89	0.12
Colon, male, 85+	0.87	0.64	0.55	0.97	0.94	0.04
Rectal, female, 20–49	0.89	0.65	0.55	1.00	0.94	0.03
Rectal, female, 50–59	0.97	0.88	0.11	0.98	0.91	0.10
Rectal, female, 60–69	0.95	0.78	0.30	0.98	0.89	0.14
Rectal, female, 70–84	0.93	0.65	0.52	0.98	0.89	0.14
Rectal, female, 85+	0.91	0.73	0.38	0.96	0.93	0.09
Rectal, male, 20–49	0.98	0.91	0.10	0.98	0.93	0.06
Rectal, male, 50–59	0.84	0.49	0.86	0.99	0.94	0.05
Rectal, male, 60–69	0.86	0.53	0.79	0.98	0.94	0.07
Rectal, male, 70–84	0.93	0.77	0.34	0.99	0.92	0.10
Rectal, male, 85+	0.97	0.92	0.03	0.98	0.92	0.07

We compare the original calibrated model (\(p_{orig}(\theta \mid y)\)), the start from scratch method for recalibration that incorporates new screen-detection rate target (\(p_{scratch}(\theta \mid y,z)\)) and the sequential recalibration method that starts from the original calibrated posterior samples when adding in the new target (\(p_{seq}(\theta \mid y,z)\))

ISSN: 1472-6947