polle (version 1.6.2)

get_policy_functions.blip: Get Policy Functions

Description

get_policy_functions() returns a function defining the policy at the given stage. get_policy_functions() is useful when implementing the learned policy.

Usage

# S3 method for blip
get_policy_functions(
  object,
  stage,
  threshold = NULL,
  include_g_values = FALSE,
  ...
)
# S3 method for drql
get_policy_functions(
  object,
  stage,
  threshold = NULL,
  include_g_values = FALSE,
  ...
)
get_policy_functions(object, stage, threshold, ...)
# S3 method for ptl
get_policy_functions(object, stage, threshold = NULL, ...)
# S3 method for ql
get_policy_functions(
  object,
  stage,
  threshold = NULL,
  include_g_values = FALSE,
  ...
)

Value

Functions with arguments:

H

data.table::data.table containing the variables needed to evaluate the policy (and g-function).

Arguments

object

Object of class "policy_object" or "policy_eval", see policy_learn and policy_eval.

stage

Integer. Stage number.

threshold

Numeric, threshold for not choosing the reference action at stage 1.

include_g_values

If TRUE, the g-values are included as an attribute.

...

Additional arguments.

Examples

library("polle")
### Two stages:
d <- sim_two_stage(5e2, seed=1)
pd <- policy_data(d,
                  action = c("A_1", "A_2"),
                  baseline = "BB",
                  covariates = list(L = c("L_1", "L_2"),
                                    C = c("C_1", "C_2")),
                  utility = c("U_1", "U_2", "U_3"))
pd

### Realistic V-restricted Policy Tree Learning
# specifying the learner:
pl <- policy_learn(type = "ptl",
                   control = control_ptl(policy_vars = list(c("C_1", "BB"),
                                                            c("L_1", "BB"))),
                   full_history = TRUE,
                   alpha = 0.05)

# evaluating the learner:
pe <- policy_eval(policy_data = pd,
                  policy_learn = pl,
                  q_models = q_glm(),
                  g_models = g_glm())

# getting the policy function at stage 2:
pf2 <- get_policy_functions(pe, stage = 2)
args(pf2)

# applying the policy function to new data:
set.seed(1)
L_1 <- rnorm(n = 10)
new_H <- data.frame(C = rnorm(n = 10),
                    L = L_1,
                    L_1 = L_1,
                    BB = "group1")
d2 <- pf2(H = new_H)
head(d2)