Table 1.

Comparison of model performance in predicting the abundance of infected individuals over a 28 day interval

Loss rate term	Loss rate =	Relative mean -log10(likelihood)		Relative DIC
		No site differences	Site differences	No site differences	Site differences
Seasonal (death + recovery)	M(1+α sin(2π(t−Δ)))I(t)	0	−1.6 (−4.4, 1.6)	0	1.3
Constant (death + recovery)	γI(t)	14 (12, 16)	12 (10, 15)	13	13
Recovery 28 days, seasonal death	$\frac{\beta S\left(\tau \right)/\left(\tau \right)\Lambda \left(t\right)}{N{\left(\tau \right)}^q}+D\left(t\right)/\left(t\right)$	37 (35, 39)	36 (33, 39)	37	38
Recovery 28 days, constant death	$\frac{\beta S\left(\tau \right)/\left(\tau \right)\exp \left(-\lambda \epsilon \right)}{N{\left(\tau \right)}^q}+\lambda /\left(t\right)$	44 (42, 46)	41 (39, 44)	44	42

M, α, and Δ are the mean, amplitude, and phase shift, respectively. γ is the constant loss rate. τ = t − ε, where the period of infection ε is 28 days. Seasonal loss rate Λ (t) = exp(−∫_τ^t D(x)Δ x), where seasonal mortality is D(t) = b(1 + a sin(2π (t − σ ))), with b, a, and σ being the mean, amplitude, and phase shift, respectively. λ is the constant death rate. 95% credibility intervals for −log₁₀(likelihood) are given in parentheses.