compute_rad: Compute risk-adjusted disparate impact estimate for a policy
In jongbinjung/undi: Test for Unjustified Disparate Impact

Compute risk-adjusted disparate impact estimate for a policy

1
2
3

compute_rad(pol, controls = NULL, base_group = NULL,
  minority_groups = NULL, fit_fn = "logit_coef", down_sample = 1,
  use_speedglm = TRUE, seed = round(stats::runif(1) * 10000), ...)

`pol`	object of class policy
`controls`	character vector of additional controls to consider in the second-stage model
`base_group`	(Optional) single group that acts as the pivot/base; by default, if the grouping variable is a factor, set to the first level, otherwise set to the first of sorted unique values
`minority_groups`	(Optional) groups to compare to the base group; by default, set to every unique value other than the base group
`fit_fn`	string indicating the rad estimation model/procedure used. `_coef` methods use models without interaction between risk and group, and return the coeficient on group membership. `_avg` methods will fit more flexible models (possibly with interactions), and compute average ratios across the population. (TODO: better documentation is expected)
`down_sample`	(Optional) proportion (between 0 and 1) or number (greater than 1) of rows to sample, if down sampling the (test) data; default is 1 (i.e., use all data)
`use_speedglm`	whether or not to use `speedglm`, instead of `stats::glm`, in cases where N > 2P (see details)
`seed`	random seed to set
`...`	additional arguments to pass to `fit` function from `rad_control` for fine-tuning

tidy data frame with columns

`term`	the group members considered minority
`estimate`	log-odds of treatment, relative to base_group (equivalent to logistic regression coefficient)
`std.error`	coefficient standard errors for `_coef` methods; TO BE IMPLEMENTED for `_avg` methods (FOR NOW, ALL ZERO)!
`statistic/p.value`	(for `*_coef` methods) corresponding values from model fit
`ptrt_base/minor`	(for `*_avg` methods) estimated average treatment probability for base/minority groups
`method`	the method used
`controls`	features controlled for