Table 7.

MANCOVA analyses examining how shape variation is related to various factors. As a result of finding significant interactions between ecomorph and genetic cluster, separate MANCOVA analyses were performed for each genetic cluster (St. Lawrence, Hudson, and Ontario). Note that partial variance explained does not sum to unity (Langerhans and DeWitt 2004).

	F	df (numerator, denominator)	P	Wilk's partial λ	Partial variance explained (%)
All Populations
Centroid size	41.33	28,445	0.0001	0.278	72.2
Ecomorph	4.53	28,445	< 0.0001	0.778	22.2
Genetic cluster	24.28	56,890	< 0.0001	0.156	60.4
Ecomorph × cluster	2.58	56,890	< 0.0001	0.740	13.9
Lake (Cluster)	7.45	252,3845	< 0.0001	0.033	26.7
St. Lawrence populations
Centroid size	8.87	28,90	< 0.0001	0.266	73.4
Ecomorph	3.24	28,90	< 0.0001	0.498	50.2
Lake	7.50	56,180	< 0.0001	0.090	70.0
Ecomorph × lake	1.48	56,180	< 0.0283	0.469	31.5
Hudson populations
Centroid size	8.11	28,89	< 0.0001	0.110	89.0
Ecomorph	2.17	28,89	0.0032	0.594	40.6
Lake	6.06	56,178	< 0.0001	0.118	65.6
Ecomorph × lake	1.18	56,178	0.214	0.533	27.0
Ontario populations
Centroid size	13.89	28,201	< 0.0001	0.341	65.9
Ecomorph	6.68	28,201	< 0.0001	0.694	30.6
Lake	7.50	140,997	< 0.0001	0.029	83.0
Ecomorph × lake	3.79	140,997	< 0.0001	0.122	65.1