Logistic regression models for predicting feeding presence/absence at GPS cluster sites were compared with AIC model selection, feeding event predictive performance, and predicted number of feeding events.

Each component includes:*log(POSCOUNT), NIGHTPROP, log(POSCOUNT)*NIGHTPROP, CENTR, CENTR2†ACCX_AVG, ACCX_AVG2, ACCXYDIFF_AVG‡SEARCH, FIELDPROP, FIELDPROP2§SUM (binary)For simplification, models shown are reduced to components based on: Spatio-temporal (SpTemp), Accelerometer activity sensor (Acvty), ground-truthing errors (GrndTru), and calendar season (Seas) covariates.