Subset Model Results for Binary and Categorical Exposures

This function extracts and combines evaluation tables for specified outcome variables from model results, handling both binary and categorical exposure models. It provides enhanced functionality for creating and analyzing subsets based on various conditions.

Usage

margot_subset_model(
  model_results,
  outcome_vars = NULL,
  subset_condition = NULL,
  scale = "RD",
  contrast = NULL,
  X = NULL,
  subset_var = NULL,
  subset_value = 1,
  subset_operator = "==",
  subset_description = NULL,
  debug = FALSE
)

Arguments

model_results: A list of model results from `model_causal_forest` or similar functions.
outcome_vars: Optional. A character vector of outcome variable names. If NULL, the function will use all models in the input.
subset_condition: A logical vector indicating the subset of data to use. Default is NULL. Can be generated using the subset_var, subset_value, and subset_operator parameters.
scale: A character string indicating the scale to use. Default is "RD".
contrast: For categorical exposures, a single character string specifying the contrast to extract.
X: Optional. The feature matrix used in the original models. Required if using subset_var.
subset_var: Optional. The name of the variable to use for subsetting.
subset_value: Optional. The value to compare against for subsetting. Default is 1.
subset_operator: Optional. The operator to use for comparison ("==", ">", ">=", "<", "<=", "!="). Default is "==".
subset_description: Optional. A description of the subset for reporting. If NULL, one will be generated.
debug: Logical. If TRUE, print debug information. Default is FALSE.

Value

A list containing:

results: A data frame combining all custom evaluation tables for the specified outcomes and contrast.
subset_condition: The logical vector used for subsetting.
subset_description: A description of the subset.
subset_info: Additional information about the subset, if available.

Details

The function can be used in two ways: 1. By providing a pre-computed logical vector as `subset_condition`. 2. By providing a variable name (`subset_var`), value (`subset_value`), and operator (`subset_operator`) to generate the subset condition automatically.

For categorical exposures, specify the contrast parameter to extract specific comparisons.

Examples

if (FALSE) { # \dontrun{
# Example 1: Basic subsetting with a pre-computed condition
subset_condition_conservative <- X[, "t0_political_conservative_z"] > 1
model_subset_conservative <- margot_subset_model(
  model_results = models_binary,
  subset_condition = subset_condition_conservative,
  debug = FALSE
)

# Example 2: Using the built-in subsetting functionality
model_subset_religious <- margot_subset_model(
  model_results = models_binary,
  X = X,
  subset_var = "t0_religion_bigger_denominations_not_rel_binary",
  subset_value = 1,
  subset_description = "Effects among religious participants",
  debug = TRUE
)

# Example 3: For categorical exposures with specific contrast
model_subset_gen_z <- margot_subset_model(
  model_results = models_cat,
  X = X,
  subset_var = "t0_gen_cohort_gen_Z_binary",
  subset_value = 1,
  contrast = "[6.0,7.0] - [1.0,5.0)",
  scale = "RD",
  debug = FALSE
)

# Example 4: Multiple outcome variables
model_subset_boomers <- margot_subset_model(
  model_results = models_binary,
  outcome_vars = c("t2_wellbeing_z", "t2_depression_z"),
  X = X,
  subset_var = "t0_gen_cohort_gen_Boomers_binary",
  subset_value = 1,
  subset_operator = "==",
  subset_description = "Effects among Baby Boomer participants",
  debug = TRUE
)

# Example 5: More complex subsetting (multiple conditions)
# Define a complex condition directly
complex_condition <- X[, "t0_political_conservative_z"] > 1 & X[, "t0_age_z"] > -2
model_subset_complex <- margot_subset_model(
  model_results = models_binary,
  subset_condition = complex_condition,
  subset_description = "Conservative (>1 SD) and not very young (>-2 SD in age)",
  debug = FALSE
)
} # }