Table 2. Results of separate linear mixed models on dependent variables and the direction of any relationship (D) using mixed models in SAS (PROC MIXED; SAS Institute, Inc., Cary, NC) and independent variables with site as a random effect and week or individual survey ^d as the repeated measure (α = 0.05) from winter (mid December 2008–February 2009) or early winter (mid December 2008–early January 2009) in eastern Mississippi, western Tennessee, and the Mississippi Alluvial Valley, USA.

Independent Variables		Dependent Variables
		Dabbling Ducks Feeding e					Proportion Dabbling Ducks Feeding
		F a	DFb b	DFw c	P	D	F a	DFb b	DFw c	P	D
Winter	Week	6.29	9	72.1	<0.001	–	2.45	9	71.5	0.017	+
	Treatment	0.40	2	25.3	0.671		0.34	2	70.9	0.715
	Water depth	8.38	1	65.1	0.005	–	4.80	1	54.3	0.033	–
Early winter	Survey d	3.12	4	33.3	0.027	–	1.46	4	32.5	0.236
	Treatment	0.85	2	32.3	0.435		0.74	2	31.2	0.485
	Water Depth	0.17	1	20	0.683		5.06	1	13.1	0.042	–

^a F-statistic from linear mixed models

^b Degrees of freedom between groups

^c Degrees of freedom within groups

^d Individual surveys (n = 5)

^e Square root transformation used to normalize distribution of residuals and homogenize variances