Skip to contents

S3 Methods for prince_bart Objects

prince_bart-class.Rd

Print, summary, and coef methods for objects returned by prince_BART. Behavior is specialized by modality: binary fits report principal strata and treatment effects for compliers, while ordinal fits report contrasts for the monotone compliance group and stratified by baseline uptake level.

Usage

# S3 method for class 'prince_bart'
print(x, ...)

# S3 method for class 'prince_bart'
summary(
  object,
  type = c("mixed", "sample"),
  treated_only = FALSE,
  adaptive_levels = TRUE,
  cumulative_mass = 0.8,
  level_threshold = NULL,
  ...
)

# S3 method for class 'prince_bart'
coef(
  object,
  type = c("mixed", "sample"),
  treated_only = FALSE,
  adaptive_levels = TRUE,
  cumulative_mass = 0.8,
  level_threshold = NULL,
  ...
)

Arguments

x, object

A prince_bart object (either prince_bart_binary or prince_bart_ordinal).

...

Additional arguments (currently ignored).

type

Character; type of estimand: "mixed" (default) for mixed estimands averaging over principal stratum uncertainty, or "sample" for sample estimands using imputed principal strata (experimental). For ordinal fits, this parameter is accepted but may be ignored depending on implementation.

treated_only

Logical; if TRUE, compute treatment effect only among the treated units (ATT). Default is FALSE (ATE). For ordinal fits, this parameter is accepted but may be ignored or handled differently.

adaptive_levels

Logical; for ordinal fits, if TRUE (default), show level-specific effects up to a data-adaptive threshold based on cumulative affected-unit mass.

cumulative_mass

Numeric in (0, 1]; for ordinal fits with adaptive grouping, levels are shown individually until this cumulative mass is reached. Default is 0.80.

level_threshold

Optional integer threshold K for ordinal fits. If supplied, levels 1..K are shown individually and higher levels are pooled as "Level > K". This overrides adaptive grouping.

Value

print: Invisibly returns the object. summary: Prints and invisibly returns a list with posterior summaries (including uncertainty and diagnostics) appropriate to the modality. coef: Named numeric vector of posterior mean estimates for the key estimands.

Examples

if (FALSE) { # \dontrun{
# Binary fit
fit_binary <- prince_BART(Y ~ X | Z | W, data = df)
summary(fit_binary)
coef(fit_binary)

# Ordinal fit
fit_ordinal <- prince_BART(Y ~ X | Z | W, data = df, uptake_type = "ordinal")
summary(fit_ordinal)
coef(fit_ordinal)
} # }