Aggregating some variables could be problematic, and may need to add a way to weight them by total population (with each given census unit) and/or number of households.
For median household income, aggregating a median value could be especially problematic (because the distribution of values changes when aggregating). May want to use the full demographic profiles / brackets.
Aggregating some variables could be problematic, and may need to add a way to weight them by total population (with each given census unit) and/or number of households.
For median household income, aggregating a median value could be especially problematic (because the distribution of values changes when aggregating). May want to use the full demographic profiles / brackets.
For example, see: https://acsdatacommunity.prb.org/discussion-forum/f/forum/542/aggregating-median-hh-income-and-calculating-moes In particular, note the link to the LA Times Data Desk's
census-data-aggregator
: https://github.com/datadesk/census-data-aggregator#approximating-mediansThat repository links to a document that describes the problem and has a proposed solution: https://www.documentcloud.org/documents/6165014-How-to-Recalculate-a-Median.html#document/p4/a508562
Also see this discussion: https://acsdatacommunity.prb.org/discussion-forum/f/forum/898/allocating-median-household-income-across-census-boundaries