Spatial analysis of bladder, kidney, and pancreatic cancer on upper Cape Cod: an application of generalized additive models to case-control data

Table 2 Summary of bladder cancer models showing degree of smoothing and global test statistic

Analysis	Latency (yrs)	Span^a	Cases/Controls^c	Deviance p-value^d	Figure #
Crude All Residences	0	0.90	95/1382	0.57	--
Adjusted All Residences	0	0.90	95/1382	0.36	3a
Adjusted All Residences	15	0.30	45/665	0.05	3b
Crude All Residences	15	0.40	45/665	0.11	--
Crude All Residences	15	0.30^b	45/665	-----^e	3c
Adjusted Longest Duration Residence	15	0.30^b	29/417	-----^e	3d
Adjusted Longest Duration Residence	15	0.95	29/417	0.41	--

^{Span is the percentage of data used for smoothing the model.}
^a Optimal span obtained by using the Akaike's Information Criterion (AIC) unless otherwise noted.
^b Same span as for adjusted, 15 years latency, all residences.
^c Number of residences contributed to the analysis by participant status.
^d Null hypothesis is that the map is flat.
^e The p-values were omitted because the maps were created using a non-optimal span.

ISSN: 1476-069X