sim_goal_weights: Simulate goal weights

In taylordunn/gasr: Simulate and Analyze Goal Attainment Scaling Data

Simulate goal weights

Description

Adds numeric weights to a set of goals (usually a single subject's set of goals).

Usage

sim_goal_weights(
  goals,
  weight_type = c("unweighted", "preference", "treatment")
)

Arguments

goals	A data frame with a row for each goal.
weight_type	The type of weights to apply. See 'Details'.

Details

There are a few valid options to the weight_type argument, including "unweighted", "preference", and "treatment". The "unweighted" type is the default and is the simplest: it sets all goal weights to 1. The "preference" type randomly randomly samples (without replacement) integer weights from 1 to n. For instance, a subject with n = 3 goals will have weights 1, 2, and 3 randomly assigned to their goals. The "treatment" type computes a weight biased by the treatment effect, and so requires a numeric treatment_fe column. These weights are calculated as n * treatment_fe / sum(treatment_fe).

Value

The same goals data frame with the added goal_weight column.

Examples

library(dplyr)
library(tidyr)
library(purrr)

# Generate some subjects and goals
d <- sim_subjects(n_subjects = 40) %>%
  sim_goals(n_goals_range = c(1, 3))

# Apply weights in one of two ways
# Using purrr::map()
d %>%
  mutate(
    goals = map(
      goals, sim_goal_weights
    )
  )
# Or using tidyr::unnest() on the goals column
d %>%
  unnest(goals) %>%
  group_by(subject_id) %>%
  sim_goal_weights(weight_type = "preference")

# The "treatment" goal weights require a treatment_fe column
d %>%
  unnest(goals) %>%
  group_by(subject_id) %>%
  sim_treatment_effect(delta = 0.4) %>%
  sim_goal_weights(weight_type = "treatment")

taylordunn/gasr documentation built on April 5, 2022, 1:37 a.m.