Package: UHDmedi
Type: Package
Title: Causal Mediation Analysis for Low- and High-Dimensional Data
Version: 0.3.0
Authors@R: person("Bo", "Shi", role = c("aut", "cre"),
                  email = "tboshi@amazon.com")
Author: Bo Shi [aut, cre]
Maintainer: Bo Shi <tboshi@amazon.com>
Description: Provides a common interface to estimators of natural indirect,
    natural direct, and total effects for low- and high-dimensional mediation
    data under method-specific structural assumptions. It includes
    a sparsity-based balancing estimator, a two-fold linear factor estimator
    based on diversified projections, and a four-fold nonlinear MediEncoder
    estimator backed by Python and PyTorch. An optional adapter to the
    'medoutcon' package provides low-dimensional natural-effect estimation
    with Z set to NULL and covariance-aware total-effect inference.
License: MIT + file LICENSE
Encoding: UTF-8
Depends: R (>= 3.6.0)
Imports:
    glmnet,
    graphics,
    grDevices,
    parallel,
    quadprog,
    splines,
    stats,
    utils
Suggests:
    medoutcon (>= 0.2.4),
    pogs,
    reticulate
Additional_repositories: https://nhejazi.r-universe.dev
SystemRequirements: Python (>= 3.9), PyTorch, NumPy, scikit-learn (for medi_encoder only)
RoxygenNote: 7.3.2
