Table 2.

Model selection results for analysis of diel variation in non-fatal bird-window collisions.

Logit model	Count model	Ka	ΔAICcb	AIC weightc
Null	SurveyTime	5	0.0	0.69
SurveyTime	Null	5	1.6	0.31

Confidence set models (i.e., those with ΔAICc ≤ 2) resulting from full model selection comparison among candidate zero-inflated Poisson regression models that included factors potentially affecting numbers of non-fatal bird collisions (including stunned birds or birds we observed to collide and then fly away) found in morning, midday, and evening collision surveys in Stillwater, Oklahoma, USA in 2015–2016. Parameters in logit and count models included Season (spring, summer, fall) and SurveyTime (survey start time in decimal format). Weights are based solely on comparison of these models; full model selection results are shown in Supplementary Table S2.

^aNumber of model parameters.

^bDifference in AICc value between model and top model.

^cAICc Weight—relative support for model.