Table 3. Generalized linear models (Gaussian error distribution) examining effects of bee diversity on fruit set, berry weight and seed set benefits due to pollination.

Full models included honey bee and wild bee abundance and Chao1 estimated bee richness for Michigan (MI) and Shannon diversity (H) for British Columbia (BC), with interaction terms for each. Among correlated explanatory variables the only the one which showed the best AICc and pseudo-R2 were used. Models with ΔAIC_c ≤ 3 are not shown.

Response variable	Region	Explanatory variables	df	logLik	AICc	ΔAICc	Model weight
fruit set gain	MI*	Honey bees	3	14.481	-20.96	0	0.57
		Honey bees + Chao1 richness	4	15.36	-19.07	1.89	0.22
	BC	[null]	2	18.75	-32.16	0	0.47
		H	3	19.81	-30.62	1.54	0.22
		Wild bees	3	19.21	-29.42	2.74	0.12
fruit weight benefit	MI	Honey bees	3	2.62	2.76	0	0.57
		Honey bees x wild bees	5	5.32	5.35	2.59	0.16
	BC	Wild bees	3	4.99	-0.97	0	0.33
		Wild bees + H	4	6.97	-0.22	0.75	0.22
		H	3	4.30	0.40	1.37	0.16
		(null)	2	2.32	0.69	1.66	0.14
seed set	MI	Honey bees	3	-49.19	106.39	0	0.58
		Honey bees + wild bees	4	-48.34	108.32	1.93	0.23
	BC	Wild bees*	3	-32.25	73.50	0	0.75

Asterisks indicate deviation from normality in the response variable (Shapiro-Wilk: W = 0.88, p = 0.038) or residuals (Shapiro-Wilk: W = 89, p = 0.048).