diff --git a/src/municipalities/02-CA.R b/src/municipalities/02-CA.R index b4f2f6c..d4d491a 100644 --- a/src/municipalities/02-CA.R +++ b/src/municipalities/02-CA.R @@ -2,26 +2,25 @@ # 02-CA.R · Correspondence Analysis of the 2022 municipal cross-section # ============================================================================= # -# Bourdieusian framework: # - Individuals (rows): 290 Swedish municipalities # - Active columns: population-composition counts — variables that all # measure "number of residents with attribute X", # sharing a common unit (persons). The active set is # deliberately restricted to age structure, -# educational capital (Swedish-born, by gender), +# educational attainment (Swedish-born, by gender), # employment sector, and national origin. These -# variables together define the compositional profile +# variables together define the profile # of the municipality's resident population. -# - col.sup (a): educational provision counts — the research object -# - col.sup (b): political vote and council counts — outcomes -# - col.sup (c): infrastructure & mobility counts — dwellings, +# - col.sup (a): educational provision counts +# - col.sup (b): political vote and council counts +# - col.sup (c): infrastructure & mobility counts, # vehicles, commuter flows, agricultural enterprises, # livestock; genuine counts but measuring different # entities (not persons), so they cannot share a # contingency table with the active columns # - Outside CA: rates, proportions, continuous measures, and # count variables measuring event flows (births, -# deaths, migration) → correlated with CA row +# deaths, migration) correlated with CA row # scores post-hoc # # Why restrict active variables to person-counts?