declare_potential_outcomes: Declare potential outcomes

Description

Declare potential outcomes

Usage

declare_potential_outcomes(..., handler = potential_outcomes_handler,
  label = NULL)
potential_outcomes.formula(formula, conditions = c(0, 1),
  assignment_variables = "Z", data, level = NULL,
  label = outcome_variable)
potential_outcomes.NULL(formula = stop("Not provided"), ..., data,
  level = NULL)

Arguments

...

arguments to be captured, and later passed to the handler

handler

a tidy-in, tidy-out function

label

a string describing the step

formula

a formula to calculate potential outcomes as functions of assignment variables.

conditions

see expand_conditions. Provide values (e.g. conditions = 1:4) for a single assignment variable. If multiple assignment variables, provide named list (e.g. conditions = list(Z1 = 0:1, Z2 = 0:1)). Defaults to 0:1 if no conditions provided.

assignment_variables

The name of the assignment variable. Generally not required as names are taken from conditions.

data

a data.frame

level

a character specifying a level of hierarchy for fabricate to calculate at

Value

a function that returns a data.frame

Details

A declare_potential_outcomes declaration returns a function. The function takes and returns a data.frame with potential outcomes columns appended. These columns describe the outcomes that each unit would express if that unit were in the corresponding treatment condition.

Declaring a potential outcomes function requires postulating a particular causal process. One can then assess how designs fare under the postulated process. Multiple processes can be considered in a single design or across design. For instance one could declare two processes that rival theories would predict.

Potential outcomes can be declared as separate variables or by using a formula. See examples below.

Examples

Run this code

# NOT RUN {
# Declare potential outcomes using default handler

# There are two ways of declaring potential outcomes:

# As separate variables

my_potential_outcomes <- declare_potential_outcomes(
  Y_Z_0 = .05,
  Y_Z_1 = .30 + .01 * age
)

# Using a formula
 my_potential_outcomes <- declare_potential_outcomes(
   Y ~ .05 + .25 * Z + .01 * age * Z)

# `conditions` defines the "range" of the potential outcomes function
 my_potential_outcomes <- declare_potential_outcomes(
   formula = Y ~ .05 + .25 * Z + .01 * age * Z,
   conditions = 1:4
 )

# Multiple assignment variables can be specified in `conditions`. For example,
# in a 2x2 factorial potential outcome:

 my_potential_outcomes <- declare_potential_outcomes(
   formula = Y ~ .05 + .25 * Z1 + .01 * age * Z2,
   conditions = list(Z1 = 0:1, Z2 = 0:1)
 )

# You can also declare potential outcomes using a custom handler

my_po_function <- function(data) {
  data$Y_treated   <- rexp(nrow(data), .2)
  data$Y_untreated <- rexp(nrow(data), .4)
  data
}

custom_potential <- declare_potential_outcomes(handler = my_po_function)

# }