Table 2. Multivariable logistic regression model for the prediction of avian influenza virus isolation results in 4,563 waterfowl sampled in northwestern Minnesota, USA, 2007 and 2008.

Variable	Coefficient	†Robust SE	Odds Ratio (95% CI)	‡P
Year
2007 (1)	Referent		NC*	<0.001
2008 (2)	1.658	0.079
Location
Thief Lake WMA (1)	Referent		NC	<0.001
Roseau River WMA (2)	−0.055	0.145
Agassiz NWR (3)	−2.270	0.101
Week (centered)	1.163	0.043	NC	<0.001
(Week)2	−0.392	0.006	NC	<0.001
Age
Adult	Referent			<0.001
Juvenile	0.996	0.181	2.7 (1.4, 3.9)
Species type
Mallard	Referent			<0.001
Other Dabblers	−1.147	0.142	0.32 (0.24, 0.42)
Divers & Wood Ducks	−2.313	0.407	0.10 (0.04, 0.22)
Location X Week
2	−1.424	0.097	NC	<0.001
3	−0.362	0.101
Location X (Week)2
2	0.376	0.009	NC	<0.001
3	0.402	0.011
Year X Location
2 2	−0.854	0.191	NC	<0.001
2 3	1.416	0.197
Year X Week
2	−2.892	0.055	NC	<0.001
Year X Location X Week
2 2	3.078	0.102	NC	<0.001
2 3	1.810	0.126
Constant	−2.674	0.160	---	<0.001

^†Robust standard error (SE) adjusted for clustering within capture sites (n = 24).

^‡P-value based on Wald chi-square statistics.

*NC – Not calculated because the odds ratio depends on the level of the interacting variables.