Generate a dataset for estimating individualized treatment rules

This function generates a simulated dataset for estimating individualized treatment rules. The outcome variable is assumed to follow equation (1) in Section S.1.1 of the supplementary material.

simulate_data(
  N = 200,
  p = 20,
  K = 4,
  J = 4,
  propensity_func,
  main_func,
  interaction_func
)

Arguments

N	the number of subjects.
p	the number of covariates.
K	the number of treatments.
J	the number of subject groups.
propensity_func	a user-defined function that calculates true propensity scores of assigning each subject to each treatment. An example of the accepted format of this function can be found in pi.true.
main_func	a user-defined function that calculates main effects of covariates on the outcome for each subject. An example of the accepted format of this function can be found in mu.true.
interaction_func	a user-defined function that calculates interaction effects of of covariates and treatments on the outcome for each subject. An example of the accepted format of this function can be found in delta.true.

Value

A matrix containing the following columns of all subjects:

• ID: IDs.

• cluster: group memberships.

• treatment: observed/assigned treatments.

• reward: values of the (continuous) outcome.

• feature_1, ..., feature_p: values of the p covariates.

• interaction_1, ..., interaction_K: values of interaction effects of covariates and each of the K treatments.

• pi_1, ..., pi_K: propensity scores of assigning each of the K treatments.

• eps: values of random errors.

• treatment_opt: optimal treatments.

• reward_opt: outcome values of the optimal treatments.

• pi_true: propensity scores of assigning the observed treatments.

• pi_opt: propensity scores of assigning the optimal treatments.

See also

pi.true for propensity_func, mu.true for main_func, and delta.true for interaction_func.

Examples

#######################################
## The simulated dataset in Section S.1.2 of
## the supplementary material for N=200

set.seed(0)
simdata200 = simulate_data(
  N = 200, p = 20, K = 4, J = 4,
  propensity_func = pi.true, # equation (2) in Section S.1.1
  main_func = mu.true, # equation (5) in Section S.1.2
  interaction_func = delta.true # equation (6) in Section S.1.2
)