Table 1. The top five GLM models applied in this study to show the effects of predictors and their interactions on the number of sheep and goats killed per attack.

Abbreviations: AIC_c, Akaike Information Criterion corrected for small sample size; ΔAIC_c, model delta; OR, odds ratio; PA distance, distance to the nearest protected area; SE, standard error; w_i, Akaike weight of the i-th model; #, model number. * The slope was set to zero as presence was used as a categorical reference for absence.

#	First predictor	Second predictor	Slope β ± SE	OR	AICc	ΔAICc	wi
1	PA distance	Dog presence	−0.06 ± 0.14	0.94	120.11	0.00	0.37
		Dog absence	3.99 ± 1.70	53.78
2	Dog presence*		0	1	123.39	3.27	0.07
	Dog absence		0.67 ± 0.29	1.96
3	Persian ethnic group*	Dog presence	0	1	123.63	3.52	0.06
	Non-Persian ethnic group	Dog presence	0.53 ± 0.38	1.69
		Dog absence	1.10 ± 0.44	3.00
4	Kurdish ethnic group*	Dog presence	0	1	123.95	3.84	0.05
	Non-Kurdish ethnic group	Dog presence	0.51 ± 0.41	1.67
		Dog absence	1.10 ± 0.46	3.00
5	Persian ethnic group*		0	1	124.59	4.48	0.04
	Non-Persian ethnic group		0.64 ± 0.38	1.90