Table 4.

Prior specification for analysis models, longitudinal data imputation and longitudinal data simulation

Model	Parameter	Value
Pairwise independent*	prior mean of treatment arm A response for each subgroup (${\mathrm{\mu }}_{\mathrm{g}}^{\left(\mathrm{A}\right)}$)	0
	prior standard deviation of treatment arm A response for each subgroup (${\mathrm{\tau }}_{\mathrm{g}}^{\left(\mathrm{A}\right)}$)	0.3
	prior mean of treatment arm B response for each subgroup (${\mathrm{\mu }}_{\mathrm{g}}^{\left(\mathrm{B}\right)}$)	0
	prior standard deviation of treatment arm B response for each subgroup (${\mathrm{\tau }}_{\mathrm{g}}^{\left(\mathrm{B}\right)}$)	0.3
Hierarchical	hyperprior mean of all response prior mean (μ0)	0
	hyperprior standard deviation of all response prior mean (σ0)	0.1
	central parameter of inverse gamma distribution as which all response prior variance is distributed (τμ)	0.1
	weight parameter of inverse gamma distribution as which all response prior variance is distributed (τn)	2
Cluster hierarchical&	DP scale (α)	2
SLR @ - imputation for longitudinal data	prior mean of coefficient (βμ)	0.8
	prior standard deviation of coefficient (βσ)	0.4
	prior mean of intercept (αμ)	0
	prior standard deviation of intercept (ασ)	0.1
	central parameter of inverse gamma distribution as which imputed response variance is distributed (λμ)	0.18
	weight parameter of inverse gamma distribution as which imputed response variance is distributed (λn)	200/400
ITP $ - simulation for longitudinal data	fraction of the final treatment arm A response variance for each subgroup (${\mathrm{\omega }}_{\mathrm{g}}^{\left(\mathrm{A}\right)}$)	0.8
	fraction of the final treatment arm B response variance for each subgroup (${\mathrm{\omega }}_{\mathrm{g}}^{\left(B\right)}$)	0.8
	shape parameter of the exponential component for treatment arm A of each subgroup (${\mathrm{k}}_{\mathrm{g}}^{\left(\mathrm{A}\right)}$)	−10
	shape parameter of the exponential component for treatment arm B of each subgroup (${\mathrm{k}}_{\mathrm{g}}^{\left(\mathrm{B}\right)}$)	−10
Public parameter	central parameter of inverse gamma distribution as which all response variance is distributed (σμ)	1
	weight parameter of inverse gamma distribution as which all response variance is distributed (σn)	1

^*:We assume all subgroups from both treatment arms have identical prior means and standard derivation.

^&:Cluster hierarchical model maintains all hyperpriors from the hierarchical model and has additional DP scale parameter.

^@:All subgroups from both treatment arms are assumed to have same priors for the coefficient and intercept of SLR except for λ_n, and λ_n is specified as 200 for four subgroups and 400 for eight subgroups.

^$:Both fraction and shape parameter are specified identically within each subgroup from both treatment arms.