ethnicolr: Predict Race and Ethnicity From Name

https://github.com/appeler/ethnicolr/workflows/test/badge.svg https://ci.appveyor.com/api/projects/status/u9fe72hn8nnhmaxt?svg=true https://img.shields.io/pypi/v/ethnicolr.svg https://anaconda.org/soodoku/ethnicolr/badges/version.svg https://pepy.tech/badge/ethnicolr

We exploit the US census data, the Florida voting registration data, and the Wikipedia data collected by Skiena and colleagues, to predict race and ethnicity based on first and last name or just the last name. The granularity at which we predict the race depends on the dataset. For instance, Skiena et al.’ Wikipedia data is at the ethnic group level, while the census data we use in the model (the raw data has additional categories of Native Americans and Bi-racial) merely categorizes between Non-Hispanic Whites, Non-Hispanic Blacks, Asians, and Hispanics.

DIME Race

Data on race of all the people in the DIME data is posted here The underlying python scripts are posted here

Caveats and Notes

If you picked a random individual with last name ‘Smith’ from the US in 2010 and asked us to guess this person’s race (measured as crudely as by the census), the best guess would be based on what is available from the aggregated Census file. It is the Bayes Optimal Solution. So what good are last name only predictive models for? A few things—if you want to impute ethnicity at a more granular level, guess the race of people in different years (than when the census was conducted if some assumptions hold), guess the race of people in different countries (again if some assumptions hold), when names are slightly different (again with some assumptions), etc. The big benefit comes from when both the first name and last name is known.

Install

We strongly recommend installing ethnicolor inside a Python virtual environment (see venv documentation)

pip install ethnicolr

Or

conda install -c soodoku ethnicolr
Notes:
  • The models are run and verified on TensorFlow 2.x using Python 3.7 and 3.8 and lower will work. TensorFlow 1.x has been deprecated.
  • If you are installing on Windows, Theano installation typically needs admin. privileges on the shell.

General API

To see the available command line options for any function, please type in <function-name> --help

# census_ln --help
usage: census_ln [-h] [-y {2000,2010}] [-o OUTPUT] -l LAST input

Appends Census columns by last name

positional arguments:
  input                 Input file

optional arguments:
  -h, --help            show this help message and exit
  -y {2000,2010}, --year {2000,2010}
                        Year of Census data (default=2000)
  -o OUTPUT, --output OUTPUT
                        Output file with Census data columns
  -l LAST, --last LAST  Name or index location of column contains the last
                        name

Examples

To append census data from 2010 to a file without column headers and the first column carries the last name, use -l 0

census_ln -y 2010 -o output-census2010.csv -l 0 input-without-header.csv

To append census data from 2010 to a file with column header in the first row, specify the column name carrying last names using the -l option, keeping the rest the same:

census_ln -y 2010 -o output-census2010.csv -l last_name input-with-header.csv

To predict race/ethnicity using Wikipedia full name model, if the input file doesn’t have any column headers, you must using -l and -f to specify the index of column carrying the last name and first name respectively (first column has index 0).

pred_wiki_name -o output-wiki-pred-race.csv -l 0 -f 1 input-without-header.csv

And to predict race/ethnicity using Wikipedia full name model for a file with column headers, you can specify the column name of last name and first name by using -l and -f flags respectively.

pred_wiki_name -o output-wiki-pred-race.csv -l last_name -f first_name input-with-header.csv

Functions

We expose 6 functions, each of which either take a pandas DataFrame or a CSV. If the CSV doesn’t have a header, we make some assumptions about where the data is:

  • census_ln(df, namecol, year=2000)
    • What it does:
      • Removes extra space
      • For names in the census file, it appends relevant data of what probability the name provided is of a certain race/ethnicity
Parameters  
  df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
  namecol : {string, list, int} string or list of the name or location of the column containing the last name
  Year : {2000, 2010}, default=2000 year of census to use
  • Output: Appends the following columns to the pandas DataFrame or CSV: pctwhite, pctblack, pctapi, pctaian, pct2prace, pcthispanic. See here for what the column names mean.

    >>> import pandas as pd
    
    >>> from ethnicolr import census_ln, pred_census_ln
    
    >>> names = [{'name': 'smith'},
    ...         {'name': 'zhang'},
    ...         {'name': 'jackson'}]
    
    >>> df = pd.DataFrame(names)
    
    >>> df
          name
    0    smith
    1    zhang
    2  jackson
    
    >>> census_ln(df, 'name')
          name pctwhite pctblack pctapi pctaian pct2prace pcthispanic
    0    smith    73.35    22.22   0.40    0.85      1.63        1.56
    1    zhang     0.61     0.09  98.16    0.02      0.96        0.16
    2  jackson    41.93    53.02   0.31    1.04      2.18        1.53
    
  • pred_census_ln(df, namecol, year=2000, num_iter=100, conf_int=1.0)

    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list, int} string or list of the name or location of the column containing the last name
      year : {2000, 2010}, default=2000 year of census to use
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race (white, black, asian, or hispanic), api (percentage chance asian), black, hispanic, white. For each race it will provide the mean, standard error, lower & upper bound of confidence interval

    (Using the same dataframe from example above)

    >>> census_ln(df, 'name')
          name pctwhite pctblack pctapi pctaian pct2prace pcthispanic
    0    smith    73.35    22.22   0.40    0.85      1.63        1.56
    1    zhang     0.61     0.09  98.16    0.02      0.96        0.16
    2  jackson    41.93    53.02   0.31    1.04      2.18        1.53
    
    >>> census_ln(df, 'name', 2010)
          name   race pctwhite pctblack pctapi pctaian pct2prace pcthispanic
    0    smith  white     70.9    23.11    0.5    0.89      2.19         2.4
    1    zhang    api     0.99     0.16  98.06    0.02      0.62        0.15
    2  jackson  black    39.89    53.04   0.39    1.06      3.12         2.5
    
    >>> pred_census_ln(df, 'name')
          name   race       api     black  hispanic     white
    0    smith  white  0.002019  0.247235  0.014485  0.736260
    1    zhang    api  0.997807  0.000149  0.000470  0.001574
    2  jackson  black  0.002797  0.528193  0.014605  0.454405
    
  • pred_wiki_ln( df, namecol, num_iter=100, conf_int=1.0)

    • What it does:
    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list, int} string or list of the name or location of the column containing the last name
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race (categorical variable — category with the highest probability), “Asian,GreaterEastAsian,EastAsian”, “Asian,GreaterEastAsian,Japanese”, “Asian,IndianSubContinent”, “GreaterAfrican,Africans”, “GreaterAfrican,Muslim”, “GreaterEuropean,British”,”GreaterEuropean,EastEuropean”, “GreaterEuropean,Jewish”,”GreaterEuropean,WestEuropean,French”, “GreaterEuropean,WestEuropean,Germanic”,”GreaterEuropean,WestEuropean,Hispanic”, “GreaterEuropean,WestEuropean,Italian”,”GreaterEuropean,WestEuropean,Nordic”. For each race it will provide the mean, standard error, lower & upper bound of confidence interval
    >>> import pandas as pd
    
    >>> names = [
    ...             {"last": "smith", "first": "john", "true_race": "GreaterEuropean,British"},
    ...             {
    ...                 "last": "zhang",
    ...                 "first": "simon",
    ...                 "true_race": "Asian,GreaterEastAsian,EastAsian",
    ...             },
    ...         ]
    >>> df = pd.DataFrame(names)
    
    >>> from ethnicolr import pred_wiki_ln, pred_wiki_name
    
    >>> odf = pred_wiki_ln(df,'last', conf_int=0.9)
    ['Asian,GreaterEastAsian,EastAsian', 'Asian,GreaterEastAsian,Japanese', 'Asian,IndianSubContinent', 'GreaterAfrican,Africans', 'GreaterAfrican,Muslim', 'GreaterEuropean,British', 'GreaterEuropean,EastEuropean', 'GreaterEuropean,Jewish', 'GreaterEuropean,WestEuropean,French', 'GreaterEuropean,WestEuropean,Germanic', 'GreaterEuropean,WestEuropean,Hispanic', 'GreaterEuropean,WestEuropean,Italian', 'GreaterEuropean,WestEuropean,Nordic']
    
    >>> odf
       last  first                         true_race  ...  GreaterEuropean,WestEuropean,Nordic_lb  GreaterEuropean,WestEuropean,Nordic_ub                              race
    0  Smith   john           GreaterEuropean,British  ...                                0.005248                                0.034912           GreaterEuropean,British
    1  Zhang  simon  Asian,GreaterEastAsian,EastAsian  ...                                0.002145                                0.068731  Asian,GreaterEastAsian,EastAsian
    
    [2 rows x 56 columns]
    
    >>> odf.iloc[0, :8]
    last                                                       Smith
    first                                                       john
    true_race                                GreaterEuropean,British
    Asian,GreaterEastAsian,EastAsian_mean                   0.015033
    Asian,GreaterEastAsian,EastAsian_std                    0.009354
    Asian,GreaterEastAsian,EastAsian_lb                     0.005062
    Asian,GreaterEastAsian,EastAsian_ub                     0.036501
    Asian,GreaterEastAsian,Japanese_mean                    0.003411
    Name: 0, dtype: object
    
  • pred_wiki_name(df, namecol, num_iter=100, conf_int=1.0)

    • What it does:
    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list} string or list of the name or location of the column containing the first name, last name, middle name, and suffix, if there. The first name and last name columns are required. If no middle name of suffix columns are there, it is assumed that there are no middle names or suffixes.
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race (categorical variable—category with the highest probability), “Asian,GreaterEastAsian,EastAsian”, “Asian,GreaterEastAsian,Japanese”, “Asian,IndianSubContinent”, “GreaterAfrican,Africans”, “GreaterAfrican,Muslim”, “GreaterEuropean,British”,”GreaterEuropean,EastEuropean”, “GreaterEuropean,Jewish”,”GreaterEuropean,WestEuropean,French”, “GreaterEuropean,WestEuropean,Germanic”,”GreaterEuropean,WestEuropean,Hispanic”, “GreaterEuropean,WestEuropean,Italian”,”GreaterEuropean,WestEuropean,Nordic”. For each race it will provide the mean, standard error, lower & upper bound of confidence interval

    (Using the same dataframe from example above)

    >>> odf = pred_wiki_name(df,'last', 'first', conf_int=0.9)
    ['Asian,GreaterEastAsian,EastAsian', 'Asian,GreaterEastAsian,Japanese', 'Asian,IndianSubContinent', 'GreaterAfrican,Africans', 'GreaterAfrican,Muslim', 'GreaterEuropean,British', 'GreaterEuropean,EastEuropean', 'GreaterEuropean,Jewish', 'GreaterEuropean,WestEuropean,French', 'GreaterEuropean,WestEuropean,Germanic', 'GreaterEuropean,WestEuropean,Hispanic', 'GreaterEuropean,WestEuropean,Italian', 'GreaterEuropean,WestEuropean,Nordic']
    
    >>> odf
       last  first                         true_race  ... GreaterEuropean,WestEuropean,Nordic_lb  GreaterEuropean,WestEuropean,Nordic_ub                              race
    0  Smith   john           GreaterEuropean,British  ...                               0.001280                                0.019053           GreaterEuropean,British
    1  Zhang  simon  Asian,GreaterEastAsian,EastAsian  ...                               0.000043                                0.001708  Asian,GreaterEastAsian,EastAsian
    
    [2 rows x 57 columns]
    
    >>> odf.iloc[0,:8]
    last                                                       Smith
    first                                                       john
    true_race                                GreaterEuropean,British
    __name                                                Smith John
    Asian,GreaterEastAsian,EastAsian_mean                   0.003466
    Asian,GreaterEastAsian,EastAsian_std                     0.00158
    Asian,GreaterEastAsian,EastAsian_lb                     0.001343
    Asian,GreaterEastAsian,EastAsian_ub                     0.006278
    Name: 0, dtype: object
    
  • pred_fl_reg_ln(df, namecol, num_iter=100, conf_int=1.0)

    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list, int} string or list of the name or location of the column containing the last name
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race (white, black, asian, or hispanic), asian (percentage chance Asian), hispanic, nh_black, nh_white. For each race it will provide the mean, standard error, lower & upper bound of confidence interval
    >>> import pandas as pd
    
    >>> names = [
    ...             {"last": "sawyer", "first": "john", "true_race": "nh_white"},
    ...             {"last": "torres", "first": "raul", "true_race": "hispanic"},
    ...         ]
    
    >>> df = pd.DataFrame(names)
    
    >>> from ethnicolr import pred_fl_reg_ln, pred_fl_reg_name, pred_fl_reg_ln_five_cat, pred_fl_reg_name_five_cat
    
    >>> odf = pred_fl_reg_ln(df, 'last', conf_int=0.9)
    ['asian', 'hispanic', 'nh_black', 'nh_white']
    
    >>> odf
       last first true_race  asian_mean  asian_std  asian_lb  asian_ub  hispanic_mean  ...  nh_black_std  nh_black_lb  nh_black_ub  nh_white_mean  nh_white_std  nh_white_lb  nh_white_ub      race
    0  Sawyer  john  nh_white    0.009093   0.003401  0.005307  0.015345       0.022212  ...      0.049869     0.116214     0.280148       0.775057      0.048677     0.692295     0.847054  nh_white
    1  Torres  raul  hispanic    0.006580   0.002218  0.003549  0.010928       0.874953  ...      0.005776     0.006483     0.020168       0.106233      0.021872     0.071461     0.149506  hispanic
    
    [2 rows x 20 columns]
    
    >>> odf.iloc[0]
    last               Sawyer
    first                john
    true_race        nh_white
    asian_mean       0.009093
    asian_std        0.003401
    asian_lb         0.005307
    asian_ub         0.015345
    hispanic_mean    0.022212
    hispanic_std     0.005254
    hispanic_lb       0.01498
    hispanic_ub      0.030114
    nh_black_mean    0.193638
    nh_black_std     0.049869
    nh_black_lb      0.116214
    nh_black_ub      0.280148
    nh_white_mean    0.775057
    nh_white_std     0.048677
    nh_white_lb      0.692295
    nh_white_ub      0.847054
    race             nh_white
    Name: 0, dtype: object
    
  • pred_fl_reg_name(df, namecol, num_iter=100, conf_int=1.0)

    • What it does:
    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list} string or list of the name or location of the column containing the first name, last name, middle name, and suffix, if there. The first name and last name columns are required. If no middle name of suffix columns are there, it is assumed that there are no middle names or suffixes.
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race (white, black, asian, or hispanic), asian (percentage chance Asian), hispanic, nh_black, nh_white. For each race it will provide the mean, standard error, lower & upper bound of confidence interval

    (Using the same dataframe from example above)

    >>> odf = pred_fl_reg_name(df, 'last', 'first', conf_int=0.9)
    ['asian', 'hispanic', 'nh_black', 'nh_white']
    
    >>> odf
       last first true_race  asian_mean  asian_std  asian_lb  asian_ub  hispanic_mean  ...  nh_black_std  nh_black_lb  nh_black_ub  nh_white_mean  nh_white_std  nh_white_lb  nh_white_ub      race
    0  Sawyer  john  nh_white     0.00142   0.000826  0.000607  0.003163       0.006704  ...      0.013759     0.009763     0.051053       0.963650      0.014840     0.936385     0.982274  nh_white
    1  Torres  raul  hispanic     0.00635   0.003612  0.002263  0.012572       0.888118  ...      0.005181     0.006091     0.021166       0.093334      0.031507     0.051153     0.156381  hispanic
    
    [2 rows x 20 columns]
    
    >>> odf.iloc[1]
    last               Torres
    first                raul
    true_race        hispanic
    asian_mean        0.00635
    asian_std        0.003612
    asian_lb         0.002263
    asian_ub         0.012572
    hispanic_mean    0.888118
    hispanic_std     0.034588
    hispanic_lb      0.817893
    hispanic_ub      0.928268
    nh_black_mean    0.012199
    nh_black_std     0.005181
    nh_black_lb      0.006091
    nh_black_ub      0.021166
    nh_white_mean    0.093334
    nh_white_std     0.031507
    nh_white_lb      0.051153
    nh_white_ub      0.156381
    race             hispanic
    Name: 1, dtype: object
    
  • pred_fl_reg_ln_five_cat(df, namecol, num_iter=100, conf_int=1.0)

    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list, int} string or list of the name or location of the column containing the last name
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race (white, black, asian, hispanic or other), asian (percentage chance Asian), hispanic, nh_black, nh_white, other. For each race it will provide the mean, standard error, lower & upper bound of confidence interval

    (Using the same dataframe from example above)

    >>> odf = pred_fl_reg_ln_five_cat(df,'last')
    ['asian', 'hispanic', 'nh_black', 'nh_white', 'other']
    
    >>> odf
       last first true_race  asian_mean  asian_std  asian_lb  asian_ub  hispanic_mean  hispanic_std  ...  nh_white_mean  nh_white_std  nh_white_lb  nh_white_ub  other_mean  other_std  other_lb  other_ub      race
    0  Sawyer  john  nh_white    0.100038   0.020539  0.073266  0.143334       0.044263      0.013077  ...       0.376639      0.048289     0.296989     0.452834    0.248466   0.021040  0.219721  0.283785  nh_white
    1  Torres  raul  hispanic    0.062390   0.021863  0.033837  0.103737       0.774414      0.043238  ...       0.030393      0.009591     0.019713     0.046483    0.117761   0.019524  0.089418  0.150615  hispanic
    
    [2 rows x 24 columns]
    
    >>> odf.iloc[0]
    last               Sawyer
    first                john
    true_race        nh_white
    asian_mean       0.100038
    asian_std        0.020539
    asian_lb         0.073266
    asian_ub         0.143334
    hispanic_mean    0.044263
    hispanic_std     0.013077
    hispanic_lb       0.02476
    hispanic_ub      0.067965
    nh_black_mean    0.230593
    nh_black_std     0.063948
    nh_black_lb      0.130577
    nh_black_ub      0.343513
    nh_white_mean    0.376639
    nh_white_std     0.048289
    nh_white_lb      0.296989
    nh_white_ub      0.452834
    other_mean       0.248466
    other_std         0.02104
    other_lb         0.219721
    other_ub         0.283785
    race             nh_white
    Name: 0, dtype: object
    
  • pred_fl_reg_name_five_cat(df, namecol, num_iter=100, conf_int=1.0)

    • What it does:
    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list} string or list of the name or location of the column containing the first name, last name, middle name, and suffix, if there. The first name and last name columns are required. If no middle name of suffix columns are there, it is assumed that there are no middle names or suffixes.
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race (white, black, asian, hispanic, or other), asian (percentage chance Asian), hispanic, nh_black, nh_white, other. For each race it will provide the mean, standard error, lower & upper bound of confidence interval

    (Using the same dataframe from example above)

    >>> odf = pred_fl_reg_name_five_cat(df, 'last','first')
    ['asian', 'hispanic', 'nh_black', 'nh_white', 'other']
    
    >>> odf
       last first true_race  asian_mean  asian_std  asian_lb  asian_ub  hispanic_mean  hispanic_std  ...  nh_white_mean  nh_white_std  nh_white_lb  nh_white_ub  other_mean  other_std  other_lb  other_ub      race
    0  Sawyer  john  nh_white    0.039310   0.011657  0.025982  0.059719       0.019737      0.005813  ...       0.650306      0.059327     0.553913     0.733201    0.192242   0.021004  0.160185  0.226063  nh_white
    1  Torres  raul  hispanic    0.020086   0.011765  0.008240  0.041741       0.899110      0.042237  ...       0.019073      0.009901     0.010166     0.040081    0.055774   0.017897  0.036245  0.088741  hispanic
    
    [2 rows x 24 columns]
    
    >>> odf.iloc[1]
    last               Torres
    first                raul
    true_race        hispanic
    asian_mean       0.020086
    asian_std        0.011765
    asian_lb          0.00824
    asian_ub         0.041741
    hispanic_mean     0.89911
    hispanic_std     0.042237
    hispanic_lb      0.823799
    hispanic_ub      0.937612
    nh_black_mean    0.005956
    nh_black_std     0.006528
    nh_black_lb      0.002686
    nh_black_ub      0.010134
    nh_white_mean    0.019073
    nh_white_std     0.009901
    nh_white_lb      0.010166
    nh_white_ub      0.040081
    other_mean       0.055774
    other_std        0.017897
    other_lb         0.036245
    other_ub         0.088741
    race             hispanic
    Name: 1, dtype: object
    
  • pred_nc_reg_name(df, namecol, num_iter=100, conf_int=1.0)

    • What it does:
    Parameters  
      df : {DataFrame, csv} Pandas dataframe of CSV file contains the names of the individual to be inferred
      namecol : {string, list} string or list of the name or location of the column containing the first name, last name, middle name, and suffix, if there. The first name and last name columns are required. If no middle name of suffix columns are there, it is assumed that there are no middle names or suffixes.
      num_iter : int, default=100 number of iterations to calculate uncertainty in model
      conf_int : float, default=1.0 confidence interval in predicted class
    • Output: Appends the following columns to the pandas DataFrame or CSV: race + ethnicity. The codebook is here. For each race it will provide the mean, standard error, lower & upper bound of confidence interval
    >>> import pandas as pd
    
    >>> names = [
    ...             {"last": "hernandez", "first": "hector", "true_race": "HL+O"},
    ...             {"last": "zhang", "first": "simon", "true_race": "NL+A"},
    ...         ]
    
    >>> df = pd.DataFrame(names)
    
    >>> from ethnicolr import pred_nc_reg_name
    
    >>> odf = pred_nc_reg_name(df, 'last','first', conf_int=0.9)
    ['HL+A', 'HL+B', 'HL+I', 'HL+M', 'HL+O', 'HL+W', 'NL+A', 'NL+B', 'NL+I', 'NL+M', 'NL+O', 'NL+W']
    
    >>> odf
          last   first true_race            __name     HL+A_mean  HL+A_std       HL+A_lb       HL+A_ub     HL+B_mean  ...  NL+O_mean  NL+O_std   NL+O_lb   NL+O_ub  NL+W_mean  NL+W_std   NL+W_lb   NL+W_ub  race
    0  hernandez  hector      HL+O  Hernandez Hector  2.727371e-13       0.0  2.727374e-13  2.727374e-13  6.542198e-04  ...   0.184513       0.0  0.184513  0.184513   0.001256       0.0  0.001256  0.001256  HL+O
    1      zhang   simon      NL+A       Zhang Simon  1.985419e-06       0.0  1.985417e-06  1.985417e-06  8.708256e-09  ...   0.003238       0.0  0.003238  0.003238   0.000154       0.0  0.000154  0.000154  NL+A
    
    [2 rows x 53 columns]
    
    >>> odf.iloc[0]
    last                hernandez
    first                  hector
    true_race                HL+O
    __name       Hernandez Hector
    HL+A_mean                 0.0
    HL+A_std                  0.0
    HL+A_lb                   0.0
    HL+A_ub                   0.0
    HL+B_mean            0.000654
    HL+B_std                  0.0
    HL+B_lb              0.000654
    HL+B_ub              0.000654
    HL+I_mean            0.000032
    HL+I_std                  0.0
    HL+I_lb              0.000032
    HL+I_ub              0.000032
    HL+M_mean            0.000541
    HL+M_std                  0.0
    HL+M_lb              0.000541
    HL+M_ub              0.000541
    HL+O_mean             0.58944
    HL+O_std                  0.0
    HL+O_lb              0.589441
    HL+O_ub              0.589441
    HL+W_mean            0.221308
    HL+W_std                  0.0
    HL+W_lb              0.221309
    HL+W_ub              0.221309
    NL+A_mean            0.000044
    NL+A_std                  0.0
    NL+A_lb              0.000044
    NL+A_ub              0.000044
    NL+B_mean            0.002199
    NL+B_std                  0.0
    NL+B_lb              0.002199
    NL+B_ub              0.002199
    NL+I_mean            0.000004
    NL+I_std                  0.0
    NL+I_lb              0.000004
    NL+I_ub              0.000004
    NL+M_mean            0.000008
    NL+M_std                  0.0
    NL+M_lb              0.000008
    NL+M_ub              0.000008
    NL+O_mean            0.184513
    NL+O_std                  0.0
    NL+O_lb              0.184513
    NL+O_ub              0.184513
    NL+W_mean            0.001256
    NL+W_std                  0.0
    NL+W_lb              0.001256
    NL+W_ub              0.001256
    race                     HL+O
    Name: 0, dtype: object
    

Application

To illustrate how the package can be used, we impute the race of the campaign contributors recorded by FEC for the years 2000 and 2010 and tally campaign contributions by race.

Data on race of all the people in the DIME data is posted here The underlying python scripts are posted here

Data

In particular, we utilize the last-name–race data from the 2000 census and 2010 census, the Wikipedia data collected by Skiena and colleagues, and the Florida voter registration data from early 2017.

Evaluation

  1. SCAN Health Plan, a Medicare Advantage plan that serves over 200,000 members throughout California used the software to better assess racial disparities of health among the people they serve. They only had racial data on about 47% of their members so used it to learn the race of the remaining 53%. On the data they had labels for, they found .9 AUC and 83% accuracy for the last name model.
  2. Evaluation on NC Data: https://github.com/appeler/nc_race_ethnicity

Authors

Suriyan Laohaprapanon, Gaurav Sood and Bashar Naji

Contributor Code of Conduct

The project welcomes contributions from everyone! In fact, it depends on it. To maintain this welcoming atmosphere, and to collaborate in a fun and productive way, we expect contributors to the project to abide by the Contributor Code of Conduct.

License

The package is released under the MIT License.