Spatial Selection and the Statistics of Neighborhoods

Working Paper

Using this module

Calculating DKLs

Please see example notebook with basic DKL calculations on US census data

Data Organization:

Data needs to be organized so that:

• Each row represents one neighborhood
• Each row has a unique ID.
• The data should be in a Pandas dataframe. The unique ID's should be the index (it is also sometimes helpful to keep the IDs as a column)
• A "City", or group, column should also be in the dataframe. Each city needs to have a unique identifier. You can check sizes of the cities by running: df.groupby('CITY_COLUMN').size()
• A set of "bin" columns. The bins should be counts of people/households in a bin subset.
• A column that contains the total population/household values for the row (this can be created by summing the bins)

Setting Up the DataFrame

1. Create a regular expression to match all the bin column names. The easiest way to do this is to add _bin to all the bin columns. Check your regex by running: df.filter(regex=bin_regex).columns
2. Create the CensusFrame:

    cf = CensusFrame(
        data=df,
        bin_regex=bin_regex,
        group_col='CITY_COLUMN',
        tot_col='TOTAL_POP_COLUMN',
    )

Getting Calcuations

3. You can run each calculation individually (see the functions in information.py). Or just run all of them (see notebook).
4. The calculations will create two new dataframes:
   • A neighborhood dataframe, accesseed at cf.nhood_df, this is a copy of the original dataframe with DKL and other values added
   • A city dataframe , accesseed at cf.city_df, this is a dataframe for all the group