Skip to contents

Multi-bias E-value table (v1: unmeasured confounding)

Source: R/margot_multi_evalue.R
margot_multi_evalue.Rd

Create a table-first sensitivity summary that reports baseline E-values (after optional Bonferroni CI adjustment) and, when specified, computes multi-bias E-values using the EValue package (confounding, selection, misclassification). For RD/continuous outcomes, estimates are mapped to the RR scale via the OLS-to-RR approximation (delta/sd) before multi-bias evaluation.

Usage

margot_multi_evalue(
  results,
  scale = c("RD", "RR"),
  intervention_type = c("exposure_shift", "ipsi"),
  delta_exposure = 1,
  sd_outcome = 1,
  biases = NULL,
  apply_bonferroni_first = TRUE,
  alpha = 0.05,
  m = NULL,
  notes = TRUE,
  include_baseline = TRUE,
  rename_multi = c("none", "friendly"),
  bound_params = NULL
)

Arguments

results

A data frame containing at least one of `E[Y(1)]-E[Y(0)]` (RD) or `E[Y(1)]/E[Y(0)]` (RR), plus columns `2.5 combined table row or small set of rows from a margot workflow.

scale

Character. "RD" (difference) or "RR" (ratio). If omitted, the function attempts to infer from the presence of `E[Y(1)]-E[Y(0)]` vs `E[Y(1)]/E[Y(0)]`.

intervention_type

Character. One of "exposure_shift" or "ipsi". Used only for interpretation notes.

delta_exposure

Numeric. Exposure contrast size for OLS-type E-values (RD). Defaults to 1. For IPSI, interpret as one policy contrast (α1 vs α0).

sd_outcome

Numeric. Outcome standard deviation used for OLS-type E-values (RD). Defaults to 1 (standardized outcomes).

biases

Either an EValue bias object created with `EValue::multi_bias()` (preferred), a single EValue `bias` object (e.g., `EValue::confounding()`), or a character vector of bias names to build with sensible defaults: accepted tokens are `"confounding"`, `"selection"`, `"misclassification_outcome"`, and `"misclassification_exposure"`.

apply_bonferroni_first

Logical; if TRUE (default), widen CIs using a Bonferroni correction at FWER `alpha` before computing E-values.

alpha

Numeric FWER level for Bonferroni. Default 0.05.

m

Optional integer for multiplicity (number of tests). If NULL, inferred from `nrow(results)`.

notes

Logical; if TRUE include a LaTeX-ready interpretation note in the output list.

Value

A list with elements: - `table`: data frame with adjusted CIs (if requested), baseline E-values (`E_Value`, `E_Val_bound`), and audit columns `alpha_fwer`, `m`, `scale`, `intervention_type`, `delta_exposure`, `sd_outcome`, and `bias_order`. For convenience, mirrored columns `E_value_point` and `E_value_bound` are also included. - `notes`: character vector (length 1) with an interpretation message when `notes = TRUE`.