Table 3. Estimates of the spread parameter b of the power law model for dispersal gradients based on temporal and spatial regression analysis of the spread of cucurbit downy mildew in the eastern United States.

Epidemic year	Maximum distance (km)	Temporal regression modela					Spatial regression modelb
		R2	b	LCL	UCL	P-value	R2	b	LCL	UCL	P-value
2008	1,959	0.97	1.96	1.79	2.21	0.0001	0.98	1.61	1.34	2.02	0.0005
2009	2,090	0.82	1.77	1.45	2.59	0.0007	0.96	1.51	0.94	3.95	0.0200
2010	2,221	0.83	3.36	2.85	4.15	0.0010	0.59	4.16	−26.69	1.93	0.0742
2011	1,957	0.76	2.20	1.70	3.55	0.0001	0.90	2.62	1.80	4.83	0.0037
2012	1,914	0.62	3.50	2.72	5.22	0.0001	0.95	2.32	1.77	3.35	0.0008
2013	1,953	0.81	1.90	1.58	2.60	0.0001	0.89	2.51	1.67	4.99	0.0050
2014	1,975	0.61	3.02	2.21	5.52	0.0010	0.78	3.75	2.16	14.01	0.0193
Overall meanc			2.53					2.64
Mean of years with good fitd			2.24 (2.91)				2.11 (3.96)

Notes.

^aThe parameter b is exponent of the inverse power law model. For temporal regression model, b was estimated as: e^τ∕(e^τ − 1), where τ is the slope of the regression of ln(distance) on time.

^bThe parameter b was estimated as the inverse of slope Ψ in the regression of velocity on distance. LCL and UCL are 95% lower and upper confidence limits of b, respectively, back-calculated from standard errors of least squares regression slopes; R² denotes the coefficient of determination.

^cDenotes the mean of estimates of b across all the seven epidemic years.

^dDenotes the mean of b across years where the data were well described (R² ≥ 0.80) by the power law model. Value in parenthesis is the mean of b across years where the data were poorly described (R² < 0.80) by the power law model.