the highest dimension scatter

District shape isn’t at all related to political impact (in aggregate) when it comes to redistricting.

While some weird districts may exist and those may be gerrymandered, the relationship is not statistically sound in aggregate: weird districts might be completely benign, in terms of the partisan advantage measures I’m considereding my dissertation & other classical measures of model influence.

Totally discounting the “expressive harms” arguments in this sphere, though.

I should add, the reason why this looks so clear from the outset is that, in this scattermatrix, the political variables are all related to one another, the geometric measures are all related to one another, but the relationships between political and geometric information are essentially square (or circular) scatters of points.

Maybe throwing in some further controls will help, but I doubt it…

imported from: yetanothergeographer