generics.md
In fairadapt: Fair Data Adaptation with Quantile Preservation

generics, rf

Code
  print(ad.rf)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = train, test.data = test)


  Adapting variables:
    y, x

  Based on protected attribute a

    AND

  Based on causal graph:
    y a x z
  y 0 0 0 0
  a 0 0 1 0
  x 1 0 0 0
  z 1 1 1 0

Code
  summary(ad.rf)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = train, test.data = test)

  Protected attribute:                 a
  Protected attribute levels:          0, 1
  Adapted variables:                   x, y

  Number of training samples:          100
  Number of test samples:              100
  Quantile method:                     rangerQuants

  Total variation (before adaptation): -0.0153
  Total variation (after adaptation):  -0.07085

generics, linear

Code
  print(ad.lin)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = train, test.data = test, quant.method = linearQuants)


  Adapting variables:
    y, x

  Based on protected attribute a

    AND

  Based on causal graph:
    y a x z
  y 0 0 0 0
  a 0 0 1 0
  x 1 0 0 0
  z 1 1 1 0

Code
  summary(ad.lin)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = train, test.data = test, quant.method = linearQuants)

  Protected attribute:                 a
  Protected attribute levels:          0, 1
  Adapted variables:                   x, y

  Number of training samples:          100
  Number of test samples:              100
  Quantile method:                     linearQuants

  Total variation (before adaptation): -0.0153
  Total variation (after adaptation):  -0.05233

generics, cts

Code
  print(ad.cts)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = cts, test.data = cts)


  Adapting variables:
    y, x

  Based on protected attribute a

    AND

  Based on causal graph:
    y a x z
  y 0 0 0 0
  a 0 0 1 0
  x 1 0 0 0
  z 1 1 1 0

Code
  summary(ad.cts)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = cts, test.data = cts)

  Protected attribute:                 a
  Protected attribute levels:          0, 1
  Adapted variables:                   x, y

  Number of training samples:          100
  Number of test samples:              100
  Quantile method:                     rangerQuants

  Total variation (before adaptation): 0.07294
  Total variation (after adaptation):  -0.02209

Code
  print(ad.disc)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = cts, test.data = NULL)


  Adapting variables:
    y, x

  Based on protected attribute a

    AND

  Based on causal graph:
    y a x z
  y 0 0 0 0
  a 0 0 1 0
  x 1 0 0 0
  z 1 1 1 0

Code
  summary(ad.disc)
Output

  Call:
  fairadapt(formula = y ~ ., prot.attr = "a", adj.mat = adj.mat, 
      train.data = cts, test.data = NULL)

  Protected attribute:                 a
  Protected attribute levels:          0, 1
  Adapted variables:                   x, y

  Number of training samples:          100
  Number of test samples:              0
  Quantile method:                     rangerQuants

  Total variation (before adaptation): 0.07294
  Total variation (after adaptation):  -0.03197