Prevalence of the pathogenic chytrid fungus, Batrachochytrium dendrobatidis, in an endangered population of northern leopard frogs, Rana pipiens

Table 2 Generalized linear models (GLM) of the prevalence of Bd in R. pipiens

#	Model Structure	df	dev	AIC
1	pcr~M + S + Y + T + L + S:T	307	163.19	189.19
2	pcr~M + S + Y + T + L + M:T + M:L + S:T + S:L + T:L + M:T:L + S:T:L	300	149.20	189.20
3	pcr~M + S + Y + T + S:T	308	165.77	189.77
4	pcr~M + S + Y + T + L + M:L + S:T	306	162.15	190.15
5	pcr~M + S + Y + T + L	310	170.36	190.36
6	pcr~M + S + Y + T	311	172.95	190.95
7	pcr~M + S + Y + T + L + M:Y + S:T	305	161.10	191.10
8	pcr~M + S + Y + T + L + M:T + S:T	305	161.11	191.11
9	pcr~M + S + Y + T + L + M:L	309	169.32	191.32
10	pcr~M + S + Y + T + L + M:T + M:L + T:L + M:T:L	304	159.35	191.35

We used GLM with binomial errors to model the prevalence of Bd in R. pipiens as a function of the tissue-sampling method (M), season (S), year (Y), stage class (T), location (L) and their two-way and three-way interactions. Of the 2728 models tested, we show the 10 models with the lowest Akaike Information Criterion (AIC) scores. Also shown are the residual degrees of freedom (df) and the residual deviance (dev) for each model.

ISSN: 1472-6785