Logistic regression models of variation in avian influenza virus (AIV) seroprevalence in adult waterfowl sampled in Alaska, USA, 1998–2010 (n = 3,588).

ak = number of parameters in model.bThe best approximating model has the lowest Akaike's Information Criterion (AIC) value and the highest model weight (ωi), relative to others in the model set.In this model suite, only the effects of species and sex were examined. Species include tundra swan (TUSW; Cygnus columbianus), cackling goose (CACG; Branta hutchinsii), Pacific black brant (BLBR; B. bernicla nigricans), greater white-fronted goose (GWFG; Anser albifrons), emperor goose (EMGO; A. canagica), northern pintail (NOPI; Anas acuta), Pacific common eider (COEI; Somateria mollissima v-nigrum), spectacled eider (SPEI; S. fischeri), Steller's eider (STEI; Polysticta stelleri), long-tailed duck (LTDU; Clangula hyemalis), and black scoter (BLSC; Melanitta nigra).