Table 3. Parameter estimates ± SEM for the best models predicting Andean weevil infestations based on local factors only (i.e. Best local), landscape factors only (i.e. Best landscape) or both local and landscape factors together (i.e. Best combined).

	Best combined	Best local	Best landscape
ΔAIC relative to best combined	0	+16.03	+18.18
w i	1.00	0.00	0.00
Perimeter/area ratio	0.501±0.149**	0.548±0.158**
Neighboring storage units	0.040±0.014**		0.048±0.015**
Neighboring current potato	−0.004±0.002**		−0.006±0.002**
Insecticide treatments January	−0.054±0.020**	−0.066±0.021**
Insecticide treatments December	−0.051±0.025*	−0.066±0.026**
Neighboring previous potato	0.008±0.004*
Clay	0.006±0.003*	0.006±0.003*
Elevation (sqrt transformed)	−0.028±0.013*
Number of hillings	0.107±0.066	0.126±0.068
Rotation 2007 [Not potato]	−0.083±0.060	−0.134±0.063*
Rotation 2006 [Not potato]	0.032±0.019
Height of first potato hilling	−0.007±0.004	−0.007±0.004
Organic matter		−0.012±0.006

Notes: Lower AIC values suggest better model performance. The Akaike weight, w _i, is interpreted as the relative probability that a given model is the best in the set. The models included control variables for observer effects (not presented in the table).

^* P≤0.05.

^** P≤0.01.