What if we were to take the multidimensional scaling coordinates from the voting analyses of Poole & Rosenthal, and treat these ideological coordinates as if they were subject to the same kinds of spatial autocorrelation processes we often find at the urban and regional scale? All it would take is to add one or more continuous numerical variables to each of the phenomena here -- legislators and/or the spatial constituencies they represent -- and then bring in the data to Luc Anselin's GeoDa software
Elvin Wyly, Sam Walker, and Emily Rosenman, November 23, 2012.