medfit provides unified S7-based infrastructure for fitting mediation models, extracting path coefficients, and performing bootstrap inference. It serves as the foundation package for the mediation analysis ecosystem, supporting RMediation and mediationverse.

• ADHD-Friendly API: Simple med() function for quick mediation analysis, quick() for instant results
• Effect Extractors: nie(), nde(), te(), pm(), paths() for extracting mediation effects
• Tidyverse Integration: tidy() and glance() methods for tibble-based workflows
• Unified Model Extraction: Extract mediation structure from various model types (lm, glm, lavaan)
• Flexible Model Fitting: Fit mediation models using different engines (GLM, with future support for mixed models)
• Robust Bootstrap Inference: Three bootstrap methods (parametric, nonparametric, plugin) with parallel processing
• Type-Safe S7 Classes: Modern object-oriented design with coef(), vcov(), confint(), nobs() methods

Install the stable version from CRAN:

install.packages("medfit")

Or install the development version from GitHub:

# install.packages("pak")
pak::pak("data-wise/medfit")

Or from the Data-Wise r-universe (pre-built binaries — no compiler needed):

install.packages(
  "medfit",
  repos = c("https://data-wise.r-universe.dev", "https://cloud.r-project.org")
)

library(medfit)

# Simulate data
set.seed(123)
n <- 200
mydata <- data.frame(X = rnorm(n))
mydata$M <- 0.5 * mydata$X + rnorm(n)
mydata$Y <- 0.3 * mydata$X + 0.4 * mydata$M + rnorm(n)

# Run mediation analysis in one line
result <- med(
  data = mydata,
  treatment = "X",
  mediator = "M",
  outcome = "Y"
)

# One-line summary
quick(result)
#> NIE = 0.19 | NDE = 0.16 | PM = 55%

# Individual effect extractors
nie(result)   # Natural Indirect Effect (a * b)
nde(result)   # Natural Direct Effect (c')
te(result)    # Total Effect (nie + nde)
pm(result)    # Proportion Mediated
paths(result) # All path coefficients (a, b, c')

library(generics)

# Tidy tibble of all effects
tidy(result)
#> # A tibble: 6 × 3
#>   term    estimate std.error
#>   <chr>      <dbl>     <dbl>
#> 1 a          0.448    0.107
#> 2 b          0.424    0.099
#> 3 c_prime    0.155    0.114
#> 4 nie        0.190       NA
#> 5 nde        0.155       NA
#> 6 te         0.345       NA

# One-row model summary
glance(result)
#> # A tibble: 1 × 6
#>     nie   nde    te    pm  nobs converged
#>   <dbl> <dbl> <dbl> <dbl> <int> <lgl>
#> 1 0.190 0.155 0.345  0.55   200 TRUE

coef(result)              # Path coefficients
coef(result, "effects")   # NIE, NDE, TE
vcov(result)              # Variance-covariance matrix
confint(result)           # 95% confidence intervals
nobs(result)              # Number of observations

# With bootstrap CI
result_boot <- med(
  data = mydata,
  treatment = "X",
  mediator = "M",
  outcome = "Y",
  boot = TRUE,
  n_boot = 1000,
  seed = 42
)

quick(result_boot)
#> NIE = 0.19 [0.08, 0.32] | NDE = 0.16 | PM = 55%

# If you already have fitted models
fit_m <- lm(M ~ X, data = mydata)
fit_y <- lm(Y ~ X + M, data = mydata)

med_data <- extract_mediation(
  fit_m,
  model_y = fit_y,
  treatment = "X",
  mediator = "M"
)

# Same extractors work
nie(med_data)
tidy(med_data)

• MediationData: Standardized container for simple mediation (X -> M -> Y)

  • Path coefficients (a, b, c')
  • Parameter estimates and variance-covariance matrix
  • Residual variances (for Gaussian models)
  • Variable names and metadata

• SerialMediationData: Container for serial mediation (X -> M1 -> M2 -> ... -> Y)

  • Supports product-of-three (2 mediators) and product-of-k (3+ mediators)
  • Flexible design compatible with lavaan extraction patterns
  • Extensible to complex mediation structures

• BootstrapResult: Container for bootstrap inference results

  • Point estimates and confidence intervals
  • Bootstrap distribution
  • Method metadata

Quick Start:

• med(): One-function mediation analysis (recommended starting point)
• quick(): One-line summary of results

Effect Extractors:

• nie(), nde(), te(): Natural indirect/direct and total effects
• pm(): Proportion mediated
• paths(): All path coefficients

Tidyverse Methods:

• tidy(): Convert results to tidy tibble
• glance(): One-row model summary

Base R Methods:

• coef(), vcov(), confint(), nobs(): Standard generics

Advanced:

• extract_mediation(): Extract from fitted lm/glm/lavaan models
• fit_mediation(): Fit with formula interface (GLM engine)
• bootstrap_mediation(): Bootstrap inference (parametric, nonparametric, plugin)

medfit is the foundation for the mediationverse ecosystem:

See Ecosystem Coordination for version compatibility and development guidelines.

Comprehensive articles are available on the package website:

• Get Started: Quick introduction to medfit
• Introduction: Detailed S7 class documentation
• Model Extraction: Extract from lm/glm/lavaan models
• Bootstrap Inference: Parametric and nonparametric bootstrap methods

Current Phase: Feature Complete (97%)

• Phase 1: Package setup
• Phase 2: S7 class architecture (MediationData, SerialMediationData, BootstrapResult)
• Phase 2.5: Comprehensive Quarto documentation
• Phase 3: Model extraction (lm/glm, lavaan)
• Phase 4: Model fitting (GLM engine)
• Phase 5: Bootstrap infrastructure (parametric, nonparametric, plugin)
• Phase 6: Generic functions (coef, vcov, confint, nobs, nie, nde, te, pm, paths, tidy, glance)
• Phase 6.5: ADHD-friendly API (med, quick)
• Phase 7: Polish & release

• Defensive Programming: checkmate for input validation, S7 validators for class integrity
• Testing: 584 tests with testthat (>90% coverage, enforced via codecov)
• CI/CD: R CMD check, lintr, coverage reporting via GitHub Actions

See planning/medfit-roadmap.md for detailed development plan.

This package is in active development. Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch from dev
3. Make your changes with tests
4. Submit a pull request to dev

Please note that the medfit project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

GPL (>= 3)

If you use medfit in your research, please cite:

Tofighi, D. (2026). medfit: Infrastructure for mediation analysis in R.
R package version 0.2.1. https://CRAN.R-project.org/package=medfit

• Package Documentation
• Development Guide
• Roadmap
• Ecosystem Strategy

• Author: Davood Tofighi
• Email: dtofighi@gmail.com
• ORCID: 0000-0001-8523-7776
• Issues: GitHub Issues