Algorithm 1 log-t family: ECME version of MDA

Step 1: Draw the degrees of freedom parameter $\nu$ from $\mathrm{P}(\nu \mid {\mathit{Y}}_{\mathrm{obs}},\mathit{B},\Psi )$.

Step 2: Sample the weights $\mathit{w}$ from $\mathrm{P}(\mathit{w}\mid {\mathit{Y}}_{\mathrm{obs}},\mathit{X},\mathit{B},\Psi ,\nu )$.

Step 3: Impute the missing response values that disrupt the monotone pattern from $\mathrm{P}({\mathit{Y}}_{\mathrm{MP},\mathrm{mis}}\mid {\mathit{Y}}_{\mathrm{obs}},\mathit{X},\mathit{B},\Psi ,\mathit{w})$.

Step 4: Update $\Psi$ and $\mathit{B}$ using (11).