Chandra and Sud (2012) only describe estimation of small area means, however I think it would be easy to implement small area totals in our package as well. I do not see an issue with adding this and I am not sure why the original paper only talks about estimating means rather than some general parameter $\theta$. I imagine we could compute:
$\hat Tj = \sum{i \in Nj} \hat y{ij}$
in the initial estimation and bootstrap process , and
Chandra and Sud (2012) only describe estimation of small area means, however I think it would be easy to implement small area totals in our package as well. I do not see an issue with adding this and I am not sure why the original paper only talks about estimating means rather than some general parameter $\theta$. I imagine we could compute:
$\hat Tj = \sum{i \in Nj} \hat y{ij}$
in the initial estimation and bootstrap process , and
$\widehat{MSE}(\hat Tj) = \frac{1}{B} \sum{b\in B} (\hat T_j^{(b)} - \tilde{T_j})^2$
to provide analogous estimates to means for totals.