Table 1. Comparison of models used to test the effect of host species, climate and disturbance on fly abundance in four species of bats in the Smithsonian Venezuela Project data set.

Model ranks	Model structure	K	R2	AIC	Δ AIC	w i
1a	Host*PC1 + Host*PC2 + Host*HPE	21	0.25	17180	0	99.96
2	Host*PC1 + Host*PC2	17	0.247	17196	16	0.04
3	Host*PC2	13	0.229	17328	148	0
4	Host*PC1	13	0.225	17356	176	0
5	Host*HPE	13	0.222	17378	198	0
6b	Host	9	0.211	17451	271	0
7	PC1 + PC2	8	0.079	18337	1157	0
8	PC1	7	0.077	18349	1169	0
9	PC2	7	0.063	18435	1255	0
10	HPE	7	0.06	18453	1273	0
11c	1	3	0	18806	1626	0

Models are ranked from the most supported (best model) to the least supported according to Akaike information criteria (AIC). Δ AIC – difference in AIC between the current and best model; w _i – model probabilities. K – number of parameters in the model. Note that the number of parameters includes k, which is the dispersion parameter of the negative binomial distribution (see Protocol S1 for details). Host – host species; PC1 and PC2 – principal components of the seven environmental variables; HPE – Human population density estimate (see Methods for details).

^aGlobal model; ^b host-species model; ^c Null model.