Table 1. Four nested Generalized Linear Models of native ant abundance.

	Native species density		+ Number of pitfall traps per site		+ P. chinensis abundance		+ Saturation of P. chinensis
	Parameter estimates	Effect test	Parameter estimates	Effect test	Parameter estimates	Effect test	Parameter estimates	Effect test
Intercept	3.62	P<0.0001	3.18	P<0.0001	3.30	P<0.0001	3.32	P<0.0001
Native species density	0.12	P<0.0001	0.11	P<0.0001	9.7×10−2	P<0.0001	8.1×10−2	P<0.0001
Number of pitfall trap per site	Not included	Not included	3.3×10−2	P<0.0001	4.3×10−2	P<0.0001	6.3×10−2	P<0.0001
P. chinensis abundance	Not included	Not included	Not included	Not included	−4.0×10−4	P<0.0001	−1.5×10−4	P<0.0001
Saturation of P. chinensis	Not included	Not included	Not included	Not included	Not included	Not included	−2.8×10−2	P<0.0001
−Log likelihood	571.26		584.4		607.8		664.6

Models are arranged according to increasing complexity, from left to right. The first model includes only the native species density. “+number of pitfall traps per site” includes the number of native species as well as the number of pitfall traps collected per site. “+P. chinensis abundance” model includes the number of native species, the number of pitfall traps per site and the total abundance of P. chinensis per site. “+Saturation of P. chinensis” model adds the effect of the number of pitfall traps collected where P. chinensis was present. Note that the effects of native species density and number of pitfall traps are positive on native species abundance, while both the effects of abundance and saturation of P. chinensis are negative on native species abundance. All more complex models are significantly better using −log likelihood ratio than the simpler model. All four models are exponentials of the form species abundance = e ^{Σ parameter i*value j}.