Package {RobustMediate}

Title:	Causal Mediation Analysis with Diagnostics and Sensitivity Analysis
Version:	0.3.0
Description:	Provides tools for causal mediation analysis with continuous treatments. Estimates natural direct and indirect effects by g-computation over a user-defined treatment grid, with bootstrap intervals that can resample whole clusters, and supports nonlinear dose-response through spline terms in the model formulas. Includes balance diagnostics based on standardized mean differences and stabilized inverse probability weights. Provides sensitivity analysis for confounding of the mediator-outcome relationship using the error correlation parameter of Imai, Keele, and Yamamoto (2010) <doi:10.1214/10-STS321>, and path-wise and indirect-effect robustness values in the partial R-squared metric of Cinelli and Hazlett (2020) <doi:10.1111/rssb.12348>, for both joint-significance and Sobel tests of the indirect effect. Additional utilities include visualization of dose-response mediation functions and formatted outputs for applied research.
License:	MIT + file LICENSE
URL:	https://github.com/subirhait/RobustMediate
BugReports:	https://github.com/subirhait/RobustMediate/issues
Encoding:	UTF-8
LazyData:	true
Depends:	R (≥ 4.1.0)
Imports:	cli, ggplot2 (≥ 3.4.0), rlang, scales, stats, broom
Suggests:	covr, dplyr, knitr, lme4, pkgdown, rmarkdown, splines, testthat (≥ 3.0.0), tidyr
Config/testthat/edition:	3
VignetteBuilder:	knitr
NeedsCompilation:	no
Packaged:	2026-10-26 23:39:39 UTC; subir
Author:	Subir Hait [aut, cre]
Maintainer:	Subir Hait <haitsubi@msu.edu>
Config/roxygen2/version:	8.1.1
Repository:	CRAN
Date/Publication:	2026-10-27 12:40:20 UTC

RobustMediate: Causal Mediation Analysis with Diagnostics and Sensitivity Analysis

Description

**RobustMediate** provides a workflow for causal mediation analysis with continuous treatments: g-computation estimates of natural direct and indirect effects, balance diagnostics, and sensitivity analysis.

Main functions include:

- **'robustmediate()'** - Fits treatment, mediator, and outcome models and stores precomputed results for downstream plotting and reporting.

- **'plot_balance()'** - Displays covariate balance before and after weighting for both the treatment and mediator pathways using standardized mean differences.

- **'plot_mediation()'** - Plots estimated natural direct effects (NDE) and natural indirect effects (NIE) over the treatment range, with pointwise uncertainty bands.

- **'sensitivity_rho()'** and **'plot_sensitivity()'** - Sensitivity of the indirect effect to confounding of the mediator-outcome relationship, using the correlation parameter 'rho' of Imai, Keele, and Yamamoto (2010).

- **'diagnose()'** - Produces a formatted diagnostic summary of balance, mediation effects, and sensitivity results.

Getting started

“'r library(RobustMediate)

data(sim_mediation)

fit <- robustmediate( X ~ Z1 + Z2 + Z3, M ~ X + Z1 + Z2 + Z3, Y ~ X + M + Z1 + Z2 + Z3, data = sim_mediation, R = 500 )

plot(fit) plot(fit, type = "balance") plot(fit, type = "sensitivity") diagnose(fit) “'

Sensitivity interpretation

'sensitivity_rho()' reports how the indirect effect changes as the correlation 'rho' between the mediator-model and outcome-model errors moves away from zero, and the value of 'rho' at which it reaches zero. It applies to linear models only and assumes the treatment is ignorable given the covariates.

Author(s)

Maintainer: Subir Hait haitsubi@msu.edu (ORCID)

Authors:

• Subir Hait haitsubi@msu.edu (ORCID)

References

Imai, K., Keele, L., & Yamamoto, T. (2010). Identification, inference and sensitivity analysis for causal mediation effects. *Statistical Science*, 25(1), 51–71.

See Also

Useful links:

• https://github.com/subirhait/RobustMediate

• Report bugs at https://github.com/subirhait/RobustMediate/issues

Adjust Both Paths for a Confounder of Given Strength

Description

Returns the adjusted path t-statistics and the adjusted Sobel statistic for a hypothetical confounder with the given partial R^2 values. Use it when the three values should differ, for example to explore a confounder that is strongly related to the mediator but only weakly to the treatment. See [sensitivity_paths()] for the definitions.

Usage

adjust_paths(x, r2_xu, r2_mu, r2_yu)

Arguments

Value

A data frame with the inputs and columns 't_a', 't_b', 'z_sobel', 'sig_joint', and 'sig_sobel'.

Examples

data(sim_mediation)
m_fit <- lm(M ~ X + Z1 + Z2 + Z3, data = sim_mediation)
y_fit <- lm(Y ~ X + M + Z1 + Z2 + Z3, data = sim_mediation)
sp <- sensitivity_paths(mediator_model = m_fit, outcome_model = y_fit,
                        treatment = "X", mediator = "M")
adjust_paths(sp, r2_xu = 0.02, r2_mu = 0.10, r2_yu = 0.05)

Coerce effects curve to a data frame

Description

Convenience accessor returning the NDE/NIE/TE curve data frame.

Usage

## S3 method for class 'robmedfit'
as.data.frame(x, ...)

Arguments

Value

A data frame with columns 'dose', 'estimand', 'estimate', 'lower', 'upper'.

Augment data with fitted values from a robmedfit object (broom-compatible)

Description

Returns the original data augmented with the balance weights, fitted mediator values, and fitted outcome values from the pathway models.

Usage

## S3 method for class 'robmedfit'
augment(x, data = NULL, ...)

Arguments

Value

The original data frame with additional columns:

'.ipw_weight'

Stabilised inverse probability weights from the treatment model (the weights used in the balance diagnostics).

'.fitted_mediator'

Predicted mediator values.

'.fitted_outcome'

Predicted outcome values.

'.resid_mediator'

Residuals from the mediator model.

'.resid_outcome'

Residuals from the outcome model.

Examples

fit <- robustmediate(
  treatment_formula = X ~ Z1 + Z2 + Z3,
  mediator_formula  = M ~ X + Z1 + Z2 + Z3,
  outcome_formula   = Y ~ X + M + Z1 + Z2 + Z3,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
aug <- augment(fit)
hist(aug$.ipw_weight)

Compare Two robmedfit Objects Side by Side

Description

Overlays the NDE/NIE/TE curves from two 'robmedfit' objects on the same panel. Useful for comparing model specifications (e.g. different covariate sets).

Usage

compare_fits(
  fit1,
  fit2,
  label1 = "Model 1",
  label2 = "Model 2",
  estimands = c("NDE", "NIE")
)

Arguments

Value

A 'ggplot2' object.

Examples

fit_a <- robustmediate(
  treatment_formula = X ~ Z1 + Z2 + Z3,
  mediator_formula  = M ~ X + Z1 + Z2 + Z3,
  outcome_formula   = Y ~ X + M + Z1 + Z2 + Z3,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
fit_b <- robustmediate(
  treatment_formula = X ~ Z1 + Z2,
  mediator_formula  = M ~ X + Z1 + Z2,
  outcome_formula   = Y ~ X + M + Z1 + Z2,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
compare_fits(fit_a, fit_b, label1 = "With Z3", label2 = "Without Z3")

Diagnose a robmedfit Object

Description

Prints a formatted diagnostics report covering balance, mediation effects, and sensitivity robustness. The output is structured so that it can be used directly (or with minimal editing) in the Results section of an applied paper. Returns the underlying results invisibly.

Usage

diagnose(x, ...)

Arguments

Value

Invisibly returns a list with elements 'balance', 'effects', and 'sensitivity'.

Curvature of an Effect Curve, with Bootstrap Intervals

Description

Describes the shape of an estimated effect curve (NIE, NDE, or TE) over the treatment grid: the second derivative of the curve at each interior dose, with percentile bootstrap intervals taken from the same bootstrap replicates as the effect estimates.

This is a description of the dose-response shape. It is not a sensitivity analysis for unmeasured confounding; for that see [sensitivity_rho()] and [sensitivity_paths()].

'sensitivity_curvature()' is deprecated; use 'effect_curvature()'.

Usage

effect_curvature(x, estimand = c("NIE", "NDE", "TE"), alpha = NULL)

sensitivity_curvature(x, estimand = c("NIE", "NDE", "TE"))

Arguments

Details

When the mediator and outcome models are linear (identity links) and the treatment and mediator enter only as main effects, every effect curve is a straight line by construction. The function then returns curvature exactly zero, with no interval, and says so in a message, instead of differencing rounding error. To let the curve bend, put a spline of the treatment in the model formulas, for example 'M ~ splines::ns(X, df = 3) + Z1'.

Otherwise the second derivative is computed by second differences on the (possibly unequally spaced) dose grid, for the point estimate and for each bootstrap replicate. Values at the two end doses are 'NA'.

The curvature is that of the fitted models, so it depends on the spline chosen. Natural splines are linear beyond their boundary knots, so with few degrees of freedom the curvature is pulled toward zero near the ends of the treatment range. The intervals reflect sampling variability only, not this modelling choice; compare a few spline settings before interpreting shape. Fits created before version 0.3.0 do not store replicate curves, so intervals are 'NA'.

Value

A data frame of class '"robmed_curvature"' with columns 'dose', 'estimate', 'lower', 'upper' (the effect curve and its interval), 'curvature', 'curv_lower', 'curv_upper', and 'ci_includes_zero' (whether the effect interval includes zero). Attributes 'estimand' and 'linear_by_construction' record the settings.

See Also

[plot_curvature()]

Examples

data(sim_mediation)
fit <- robustmediate(
  X ~ Z1 + Z2, M ~ splines::ns(X, df = 3) + Z1 + Z2, Y ~ X + M + Z1 + Z2,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
curv <- effect_curvature(fit, estimand = "NIE")
plot_curvature(curv, ref_dose = fit$meta$ref_dose)

Glance at a robmedfit object (broom-compatible)

Description

Returns a one-row summary of the fit: sample size, bootstrap reps, reference dose, percentage mediated, the value of 'rho' at which the indirect effect is zero ('NA' when the sensitivity analysis does not apply), and maximum post-weighting SMDs.

Usage

## S3 method for class 'robmedfit'
glance(x, ...)

Arguments

Value

A one-row data frame.

Examples

fit <- robustmediate(
  treatment_formula = X ~ Z1 + Z2 + Z3,
  mediator_formula  = M ~ X + Z1 + Z2 + Z3,
  outcome_formula   = Y ~ X + M + Z1 + Z2 + Z3,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
glance(fit)

Plot a robmedfit object

Description

Dispatches to the appropriate plot function based on 'type'.

Usage

## S3 method for class 'robmedfit'
plot(x, type = c("mediation", "balance", "sensitivity", "curvature"), ...)

Arguments

Value

A 'ggplot2' object.

Love Plot: Balance Diagnostics for Both Pathways

Description

Produces a love plot of standardised mean differences (SMDs) before and after weighting by the stabilised treatment-model weights. For the treatment pathway, covariates are compared between units above and below the median treatment. For the mediator pathway, they are compared between units above and below their predicted mediator value; this is a rough screening check, not a formal balance criterion for the mediator model.

Usage

plot_balance(x, threshold = 0.1, pathways = c("treatment", "mediator"), ...)

Arguments

Value

A 'ggplot2' object. Add layers or themes as usual.

Examples

fit <- robustmediate(
  treatment_formula = X ~ Z1 + Z2 + Z3,
  mediator_formula  = M ~ X + Z1 + Z2 + Z3,
  outcome_formula   = Y ~ X + M + Z1 + Z2 + Z3,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
plot_balance(fit)
plot_balance(fit, threshold = 0.05, pathways = "treatment")

Plot an Effect Curve and Its Curvature

Description

Two panels: the effect curve with its bootstrap interval, and its second derivative with a bootstrap interval. Doses where the effect interval includes zero are shaded.

Usage

plot_curvature(x, estimand = NULL, ref_dose = NULL, ...)

Arguments

Value

A 'ggplot2' object.

Examples

data(sim_mediation)
fit <- robustmediate(
  X ~ Z1 + Z2, M ~ splines::ns(X, df = 3) + Z1 + Z2, Y ~ X + M + Z1 + Z2,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
plot_curvature(effect_curvature(fit), ref_dose = fit$meta$ref_dose)

Dose-Response Curve: Natural Direct and Indirect Effects

Description

Plots NDE, NIE, and (optionally) total effect as smooth spline curves over the full range of treatment values, with pointwise bootstrap confidence bands. This is the signature visualisation of **RobustMediate** and is publication-ready out of the box.

Usage

plot_mediation(
  x,
  estimands = c("NDE", "NIE"),
  show_total = FALSE,
  facet = FALSE,
  ...
)

Arguments

Value

A 'ggplot2' object.

Examples

fit <- robustmediate(
  treatment_formula = X ~ Z1 + Z2 + Z3,
  mediator_formula  = M ~ X + Z1 + Z2 + Z3,
  outcome_formula   = Y ~ X + M + Z1 + Z2 + Z3,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
plot_mediation(fit)
plot_mediation(fit, estimands = c("NDE", "NIE", "TE"), facet = TRUE)

Plot Adjusted Path and Sobel Statistics Against Confounder Strength

Description

Plots the adjusted t-statistics of the two paths and the adjusted Sobel statistic as the confounder strength \rho grows (all three partial R^2 values equal to \rho). Dashed vertical lines mark the robustness values for the joint test and the Sobel test.

Usage

plot_paths(x, ...)

Arguments

Value

A 'ggplot2' object.

Examples

data(sim_mediation)
m_fit <- lm(M ~ X + Z1 + Z2 + Z3, data = sim_mediation)
y_fit <- lm(Y ~ X + M + Z1 + Z2 + Z3, data = sim_mediation)
plot_paths(sensitivity_paths(mediator_model = m_fit, outcome_model = y_fit,
                             treatment = "X", mediator = "M"))

Plot the Indirect Effect as a Function of rho

Description

Plots the indirect effect from [sensitivity_rho()] against the sensitivity parameter 'rho', with the point where the indirect effect is zero marked.

Usage

plot_sensitivity(x, ...)

Arguments

Value

A 'ggplot2' object.

Examples

data(sim_mediation)
fit <- robustmediate(
  X ~ Z1 + Z2, M ~ X + Z1 + Z2, Y ~ X + M + Z1 + Z2,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
plot_sensitivity(fit)

Print a robmed_paths object

Description

Print a robmed_paths object

Usage

## S3 method for class 'robmed_paths'
print(x, digits = 3, ...)

Arguments

Value

The input object, invisibly.

Print a robmed_sens object

Description

Print a robmed_sens object

Usage

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

Arguments

Value

The input object, invisibly.

Print a robmedfit object

Description

Print a robmedfit object

Usage

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

Arguments

Value

The input object, invisibly.

Objects exported from other packages

Description

These objects are imported from other packages. Follow the links below to see their documentation.

broom

augment, glance, tidy

Robust Causal Mediation Analysis

Description

Fits treatment, mediator, and outcome models for causal mediation analysis with continuous treatments and returns a precomputed 'robmedfit' object for plotting and diagnostics.

Natural direct and indirect effects are estimated by g-computation from the fitted mediator and outcome models, averaging over the observed covariates. When the outcome model is linear in the mediator (identity link, mediator entering as a linear term), the predicted mediator mean is plugged in, which is exact. Otherwise the outcome prediction is averaged over the mediator's distribution: exactly for a binary mediator, and by a 15-point quantile rule for a Gaussian mediator. Stabilised inverse probability weights from the treatment model are used for the balance diagnostics only.

Usage

robustmediate(
  treatment_formula,
  mediator_formula,
  outcome_formula,
  data,
  ref_dose = NULL,
  dose_grid = NULL,
  R = 500,
  alpha = 0.05,
  covariates = NULL,
  cluster_var = NULL,
  family_treatment = stats::gaussian(),
  family_mediator = stats::gaussian(),
  family_outcome = stats::gaussian(),
  spline_df = NULL,
  rho_seq = seq(-0.9, 0.9, by = 0.01),
  verbose = TRUE
)

Arguments

Value

An object of class '"robmedfit"' containing:

'models'

Fitted treatment, mediator, and outcome models.

'balance'

Balance statistics before and after weighting.

'effects'

Dose-response summaries for NDE, NIE, and TE, including bootstrap intervals.

'sensitivity'

A 'robmed_sens' object from [sensitivity_rho()], or 'NULL' when the models are not linear (the closed-form result does not apply).

'cluster'

Cluster variable, number of clusters, and cluster sizes, or 'NULL' if clustering was not used.

'meta'

Call, variable names, dose settings, bootstrap settings, and sample information.

Examples

n <- 400
Z1 <- rnorm(n)
Z2 <- rbinom(n, 1, 0.5)
X  <- 0.5 * Z1 + 0.3 * Z2 + rnorm(n)
M  <- 0.4 * X + 0.2 * Z1 + rnorm(n)
Y  <- 0.3 * X + 0.5 * M + 0.1 * Z1 + rnorm(n)
dat <- data.frame(Y, X, M, Z1, Z2)

fit <- robustmediate(
  treatment_formula = X ~ Z1 + Z2,
  mediator_formula = M ~ X + Z1 + Z2,
  outcome_formula = Y ~ X + M + Z1 + Z2,
  data = dat,
  R = 10, verbose = FALSE
)

print(fit)

Path-Wise and Indirect-Effect Robustness Values

Description

Sensitivity analysis for the two paths of a linear mediation model and for the indirect effect, in the partial R^2 metric of Cinelli and Hazlett (2020).

One unmeasured confounder 'U' is described by three partial R^2 values, all conditional on the observed covariates 'C':

• 'r2_xu': R^2_{X \sim U \mid C}, with the treatment;

• 'r2_mu': R^2_{M \sim U \mid X, C}, with the mediator;

• 'r2_yu': R^2_{Y \sim U \mid M, X, C}, with the outcome.

The a-path (treatment to mediator) depends on 'r2_xu' and 'r2_mu'. The b-path (mediator to outcome) depends on 'r2_mu' and 'r2_yu'. Because 'r2_mu' enters both paths, one confounder related to the mediator weakens both paths at once.

A robustness value (RV) is the smallest \rho such that a confounder with every partial R^2 equal to \rho changes the conclusion. The function reports:

Path RVs

'rv_q1' reduces the path estimate to zero; 'rv_alpha' makes the path test non-significant (Cinelli and Hazlett, 2020).

Indirect, joint test

The indirect effect is declared significant when both paths are. Its RV equals the smaller of the two path RVs.

Indirect, Sobel test

The indirect effect is tested by z = t_a t_b / \sqrt{t_a^2 + t_b^2}. Its RV is smaller than the smaller path RV, because one confounder lowers both t_a and t_b.

The ratio 'overstatement' = RV(joint) / RV(Sobel) shows how much the weaker-path value overstates robustness when the indirect effect is tested as a product.

Usage

sensitivity_paths(
  x = NULL,
  alpha = 0.05,
  mediator_model = NULL,
  outcome_model = NULL,
  treatment = NULL,
  mediator = NULL,
  rho_grid = seq(0, 0.6, by = 0.005)
)

Arguments

Details

The adjusted t-statistic for a coefficient with t-statistic t and residual degrees of freedom df, given treatment-side and outcome-side partial R^2 values R^2_D and R^2_Y, is

t_{adj} = \frac{|t| - \sqrt{df}\sqrt{R^2_Y R^2_D / (1 - R^2_D)}} {\sqrt{(1 - R^2_Y)/(1 - R^2_D)}\sqrt{df/(df - 1)}}

assuming the confounder biases the estimate toward zero. Path tests use the t critical value with df - 1 degrees of freedom; the Sobel test uses the normal critical value.

The result applies to linear (Gaussian, identity-link) mediator and outcome models with the treatment and mediator entering as main effects.

Value

An object of class '"robmed_paths"': a list with

'table'

Data frame with one row per path and per indirect-effect test: 'estimate', 'se', 'stat', 'df', 'rv_q1', 'rv_alpha'.

'overstatement'

RV(joint) / RV(Sobel).

'curve'

Data frame of adjusted 't_a', 't_b', and Sobel 'z' over 'rho_grid'.

'inputs'

t-statistics, degrees of freedom, and critical values.

References

Cinelli, C., & Hazlett, C. (2020). Making sense of sensitivity: Extending omitted variable bias. *Journal of the Royal Statistical Society: Series B*, 82(1), 39–67. doi:10.1111/rssb.12348

Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. *Sociological Methodology*, 13, 290–312.

See Also

[adjust_paths()], [plot_paths()], [sensitivity_rho()]

Examples

data(sim_mediation)
m_fit <- lm(M ~ X + Z1 + Z2 + Z3, data = sim_mediation)
y_fit <- lm(Y ~ X + M + Z1 + Z2 + Z3, data = sim_mediation)
sp <- sensitivity_paths(mediator_model = m_fit, outcome_model = y_fit,
                        treatment = "X", mediator = "M")
sp
plot_paths(sp)

Sensitivity of the Indirect Effect to Mediator-Outcome Confounding

Description

Implements the sensitivity analysis of Imai, Keele, and Yamamoto (2010) for the linear structural equation model. The sensitivity parameter 'rho' is the correlation between the error terms of the mediator model and the outcome model. Sequential ignorability implies 'rho = 0'. A nonzero 'rho' arises when an unmeasured pre-treatment variable affects both the mediator and the outcome.

For each value of 'rho', the indirect effect per unit change in treatment is

\delta(\rho) = \frac{\beta_2 \sigma_1}{\sigma_2} \left\{ \tilde\rho - \rho \sqrt{\frac{1 - \tilde\rho^2}{1 - \rho^2}} \right\}

where \beta_2 is the treatment coefficient in the mediator model, \sigma_2 is the residual standard deviation of the mediator model, \sigma_1 is the residual standard deviation of the outcome model with the mediator removed, and \tilde\rho is the correlation between those two sets of residuals. At 'rho = 0' this equals the product of the treatment coefficient in the mediator model and the mediator coefficient in the outcome model. The indirect effect is zero exactly when 'rho' equals \tilde\rho.

The same threshold can be stated in variance-explained terms. If an unmeasured confounder explains a proportion R^{2*}_M of the residual variance of the mediator and R^{2*}_Y of the residual variance of the outcome, then \rho^2 = R^{2*}_M R^{2*}_Y.

Usage

sensitivity_rho(x, rho_seq = seq(-0.9, 0.9, by = 0.01), contrast = NULL)

Arguments

Details

The closed form holds only for the linear model. The function therefore requires a Gaussian identity-link mediator and outcome model, a treatment that enters both models as a single linear term, and no treatment-by-mediator interaction. It treats the treatment assignment as ignorable given the covariates; it addresses only confounding of the mediator-outcome relationship. Results are point estimates; they do not include sampling uncertainty.

Value

An object of class '"robmed_sens"': a list with

'curve'

Data frame with columns 'rho', 'nie_unit' (indirect effect per unit of treatment), and 'nie' (indirect effect for 'contrast').

'rho_zero'

Value of 'rho' at which the indirect effect is zero.

'r2_product_zero'

R^{2*}_M R^{2*}_Y at which the indirect effect is zero (equal to 'rho_zero^2').

'nie_unit_0', 'contrast'

Indirect effect per unit at 'rho = 0', and the contrast used.

References

Imai, K., Keele, L., & Yamamoto, T. (2010). Identification, inference and sensitivity analysis for causal mediation effects. *Statistical Science*, 25(1), 51–71.

Imai, K., Keele, L., & Tingley, D. (2010). A general approach to causal mediation analysis. *Psychological Methods*, 15(4), 309–334.

Examples

data(sim_mediation)
fit <- robustmediate(
  X ~ Z1 + Z2, M ~ X + Z1 + Z2, Y ~ X + M + Z1 + Z2,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
sens <- sensitivity_rho(fit)
sens
plot_sensitivity(fit)

Simulated Mediation Study Data

Description

A synthetic dataset mimicking a clustered education study with a continuous treatment (tutoring hours), a continuous mediator (mid-year test score), and a continuous outcome (end-of-year test score). Designed to illustrate **RobustMediate** with realistic effect sizes and non-trivial confounding.

Usage

sim_mediation

Format

A data frame with 600 rows (30 schools x 20 students) and 7 columns:

school

Factor. School identifier (30 levels). Use as 'cluster_var'.

Y

Numeric. End-of-year test score (outcome).

X

Numeric. Tutoring hours received (continuous treatment, >= 0).

M

Numeric. Mid-year test score (mediator).

Z1

Numeric. Prior achievement (continuous covariate).

Z2

Integer (0/1). Free-lunch status (binary covariate).

Z3

Numeric. Parental education index (continuous covariate).

True parameter targets

The data-generating process sets:

• **NDE** (X → Y direct path) ~= **0.25**

• **NIE** (X → M → Y path) ~= **0.35**

• **TE** ~= **0.60**

• **

Use these as a ground truth to assess estimation accuracy.

Source

Generated via 'data-raw/generate_sim_data.R'. See that script for the full data-generating process.

Examples

data(sim_mediation)
str(sim_mediation)
summary(sim_mediation[, c("Y","X","M")])


fit <- robustmediate(
  treatment_formula = X ~ Z1 + Z2 + Z3,
  mediator_formula  = M ~ X + Z1 + Z2 + Z3,
  outcome_formula   = Y ~ X + M + Z1 + Z2 + Z3,
  data        = sim_mediation,
  cluster_var = "school",
  R           = 10,
  verbose     = FALSE
)
diagnose(fit)

Summary method for robmedfit objects

Description

Summary method for robmedfit objects

Usage

## S3 method for class 'robmedfit'
summary(object, ...)

Arguments

Value

A list with effect, balance, and sensitivity summaries, invisibly.

Tidy a robmedfit object (broom-compatible)

Description

Returns a tidy data frame of the mediation effect estimates (NDE, NIE, TE) at the reference dose, with confidence intervals. Compatible with 'broom::tidy()' and the broader 'tidymodels' ecosystem.

Usage

## S3 method for class 'robmedfit'
tidy(x, conf.int = TRUE, ...)

Arguments

Value

A data frame with columns 'term', 'estimate', 'conf.low', 'conf.high', and 'ref_dose'.

Examples

fit <- robustmediate(
  treatment_formula = X ~ Z1 + Z2 + Z3,
  mediator_formula  = M ~ X + Z1 + Z2 + Z3,
  outcome_formula   = Y ~ X + M + Z1 + Z2 + Z3,
  data = sim_mediation, R = 10, verbose = FALSE,
  dose_grid = seq(-2, 2, length.out = 25)
)
tidy(fit)