Table 2. Prior distributions of the Bayesian mixed-effects model.

parameter	description	prior	hyper-prior
σ	VL measurement error	$\left|\mathcal{N}\right(0,0.5\left)\right|$	-
ϵg	random effect VL growth rate (g)	$\mathcal{N}({\mu }_g,{\sigma }_g)$	${\mu }_g\sim \mathcal{N}(0,1)$,	${\sigma }_g\sim \left|\mathcal{N}(0,1)\right|$
αg	fixed effect of tART on g	$\mathcal{N}(0,1)$	-
ϵλ	random effect recrudescence rate (λ)	$\mathcal{N}({\mu }_{\mathrm{\lambda }},{\sigma }_{\mathrm{\lambda }})$	${\mu }_{\mathrm{\lambda }}\sim \mathcal{N}(0,1)$,	${\sigma }_{\mathrm{\lambda }}\sim \left|\mathcal{N}(0,1)\right|$
αλ	fixed effect of tART on λ	$\mathcal{N}(0,1)$	-
log10(K)	carrying capacity VL	$\mathcal{N}({\mu }_K,{\sigma }_K)$	${\mu }_K\sim \mathcal{N}(5,2)$,	${\sigma }_K\sim \left|\mathcal{N}(0,2)\right|$
log(v0)	initial VL equivalent	$\mathcal{N}(-1,1)$	-

The notation x ∼ |D| for probability distribution D means that x is positive and that D is truncated at zero. The normal distribution is parameterized with the mean and standard deviation. The time t_ART denotes the number of days post infection at which antiretroviral treatment was initiated.