How does Poisson kriging compare to the popular BYM model for mapping disease risks?

Table 1 Summary statistics for estimates of lung and cervix cancer mortality.

Estimators	Lung cancer			Cervix cancer
	Mean	Variance	Min-max	Mean	Variance	Min-Max
Observed rates	21.19	18.48	9.071–31.79	2.851	2.446	0.000–8.138
ATA Poisson kriging (PK) estimate	21.51	11.31	13.07–31.55	2.880	0.320	1.800–4.044
Point Poisson kriging (PK) estimate	21.52	11.13	13.33–31.61	2.897	0.338	1.811–4.061
BYM model estimate	21.38	11.35	13.96–31.63	2.889	0.200	1.915–4.074
ATA Poisson kriging (PK) variance	3.122	2.009	0.238–7.582	0.187	0.040	0.003–0.881
Point Poisson kriging (PK) variance	3.275	2.367	0.242–7.731	0.211	0.043	0.003–0.948
BYM model variance	3.224	3.288	0.306–12.05	0.154	0.009	0.004–0.509
Correlation PK vs ATA PK estimate	0.997			0.989
Correlation BYM vs ATA PK estimate	0.988			0.845
Correlation PK vs ATA PK variance	0.989			0.986
Correlation BYM vs ATA PK variance	0.908			0.549

Mean, variance and range of cancer mortality rates per 100,000 person-years estimated at the county level using Poisson kriging (point and ATA) and BYM model. Statistics for the kriging variance and variance of Bayesian empirical distribution are listed below.

ISSN: 1476-072X