Calibration of synthjoint output

[kuriwaki]

We want to employ some calibration into synthetic joints.

In our current calibration, we have $G \times J$ (race x district) estimates and we supply the calibration with $G + J$ constraints.

In synthjoint, we have $S_1\times S_2 \times ... S_K = S$ cells and we have some microdata on the $K$ variables to estimate the joint distribution of the covariates. But suppose the microdata is biased on to P(turnout | race) due to unignorable bias. Here say $S$ represents the covariates in a race.

Suppose an external dataset tells you the turnout rate for each race (it might even tell us turnout for each race x each of the 4 US regions or 50 states). We already know the turnout rate within a district $j$ and had incorporated it in the bmlogit. The question is, how can we incorporate an additional constraint about also the turnout rate of each race (say in the nation or region)?

@sou412