Developing the atlas of cancer in Queensland: methodological issues

Table 4 Sensitivity analyses for oesophageal cancer survival among males

	Prior 1	Prior 2	Prior 3	Prior 4	Prior 5	Prior 6
Distribution of RER
Mean	100.2	100.7	100.4	100.1	100.4	100.0
Standard deviation	6.5	11.5	8.2	3.8	9.3	0.3
Maximum	119.6	140.7	127.6	111.3	129.5	102.1
75% Quartile	105.3	105.0	105.7	102.6	106.3	100.2
Median	98.0	97.3	97.7	99.2	97.0	100.0
25% Quartile	95.2	92.6	94.7	97.2	93.7	99.8
Minimum	80.9	63.4	75.0	89.4	72.5	98.3
90% ratio¹	1.2	1.4	1.3	1.1	1.3	1.0
pD²	23.988	36.021	33.105	18.663	30.524	18.218
DIC³	3690.23	3690.27	3691.24	3691.32	3690.07	3694.96
Spatial fraction⁴	0.62	0.87	0.51	0.48	0.80	0.00
Percent SLAs with Geweke <0.01 for RER	89.3%	9.8%	10.5%	19.5%	21.5%	63.0%

Notes:
1. The 90% ratio is calculated as the 95^th percentile divided by the 5^th percentile of the smoothed RER estimates.
2. pD represents the effective number of parameters in the model. Larger values indicate less smoothing of estimates.
3. DIC = Deviance Information Criterion. Smaller values (of at least 5 below) indicate a better model fit.
4. The spatial fraction estimates the relative contribution of spatial and unstructured heterogeneity, and is calculated as: $Spatial fraction = \frac{θ_{m a r g i n a l}^{2}}{θ_{m a r g i n a l}^{2} + σ^{2}}$
Where $θ_{m a r g i n a l}^{2}$ = marginal spatial variance, σ ²= marginal variability of the unstructured random effects between areas. A value close to 1 indicates the spatial heterogeneity dominates, whereas a value close to 0 indicates the unstructured heterogeneity dominates.

ISSN: 1476-072X