Associations between ultraviolet radiation, tree cover and adolescent sunburns

Table 4 Model fit, residual spatial dependence, and residual spatial autocorrelation comparisons

	AIC	Area Under Curve (AUC)	p¹	Global Moran’s I²	p³	Count (%) of participants in significant, positive Local Moran’s I clusters
						Positive⁴	Negative⁵	Total
Model 1	1603.897	0.736	Ref	0.0033	0.71	27 (2.0%)	58 (4.4%)	85 (6.4%)
Model 2	1600.689	0.742	0.14	-0.0028	0.85	25 (1.9%)	31 (2.3%)	56 (4.2%)
Model 3	1602.285	0.742	0.14	-0.0017	0.93	30 (2.3%)	42 (3.2%)	72 (5.4%)
Model 4	1601.381	0.744	0.09	-0.0033	0.81	28 (2.1%)	29 (2.2%)	57 (4.3%)
Model 5	1602.770	0.743	0.09	-0.0023	0.89	27 (2.0%)	34 (2.6%)	61 (4.6%)

¹p-values are from Chi-square tests for differences in model Receiver Operator Curves (ROC) with Model 1 ROC
²Global Moran’s I test statistic based on the standardized deviance residuals at home addresses
³p-values are from z-scores based on a standard normal distribution of standardized deviance residuals
⁴Positive clusters represent significant (p < 0.05), positive clusters of home addresses with positive standardized deviance residuals (where the observed sunburn was higher than predicted)
⁵Negative clusters were significant (p < 0.05), positive clusters of home addresses with negative standardized deviance residuals (where the observed sunburn was lower than predicted)

ISSN: 1476-072X