NEWS

---

ipd 0.4.1.9000

ipd 0.4.0 (2026-03-06)

Summary:

• Downgrading dependence on BiocGenerics to prevent error on CRAN install.

Summary:

• Refactored internal Chen-Chen and PDC functions to centralize matrix-based estimators and reduce duplicated logic across models.

• Added thin ipd-compatible wrappers for OLS, logistic, and Poisson Chen-Chen and PDC methods.

• Improved internal GLM handling and documentation to support ongoing methodological development.

Specific Changes:

• Chen / PDC Refactor

  • Added new internal utilities file utils_chen.R housing shared score/Hessian computation and matrix-based estimators.

  • Centralized core augmented estimators:

    • compute_min_mse_est(): Chen-Chen min-MSE form
    • compute_pdc_est(): PDC augmentation
    • compute_min_mse_est_new(): experimental Chen-Chen variant with separate auxiliary design/family

  • Added shared internal helpers:

    • compute_residuals(): GLM residual and Hessian computation
    • .fit_glm_from_matrix(): GLM fitting directly from design matrices
    • .resolve_family(): robust family resolution for Gaussian, Binomial, and Poisson models
    • expit() and expit_derivative()

• Thin Wrapper Functions (ipd helpers)

  • Rewrote chen_ols() to use centralized matrix-based estimators.

  • Added thin wrappers preserving existing ipd() method lookup:

    • chen_logistic()
    • chen_poisson()
    • pdc_ols()
    • pdc_logistic()
    • pdc_poisson()

  • All wrappers maintain backward compatibility with the existing helper contract (list(est, se)).

• Internal Documentation

  • Added Roxygen2 documentation to all Chen/PDC utilities (including non-exported helpers).
  • Introduced consistent section headers for estimator components and wrappers.
  • Adding PDC method examples to documentation and vignette.
  • Updating urls throughout to use doi.org links instead of journal-specific links.

• Development Notes

  • compute_min_mse_est_new() is currently internal-only and not yet wired into the ipd() method registry; future work will determine exposure via a new method (e.g., chen_new) or argument toggle.

ipd 0.2.0

Summary:

• Preparations to archive on CRAN and move to Bioconductor.

• Slight formatting changes to conform to styler and lintr suggestions.

• Added PPIa, Chen and Chen methods from Gronsbell et al. (2025) "Another look at inference after prediction."

Specific Changes:

• DESCRIPTION updates

  • Depends: R (>= 4.4.0)
  • Added biocViews: Software
  • Added Suggests: BiocStyle, BiocManager

• Vignettes

  • Converted existing R Markdown vignettes to Bioconductor style with BiocStyle::html_document and proper VignetteIndexEntry headers.
  • Added examples for Chen and Chen, PPI "All" methods.

• NEWS & CITATION

  • Added NEWS.md entry (this file) and a CITATION file for package citation metadata.

• Testing & QA

  • Passed BiocCheck with no errors or warnings.
  • Updated testthat suite as needed for Bioc compliance.

• Continuous Integration

  • Added GitHub Actions via usethis::use_github_action("bioc-workflow") to run Bioconductor checks on Linux, macOS, and Windows.

• README

  • Installation instructions updated to:

    if (!requireNamespace("BiocManager", quietly=TRUE))
      install.packages("BiocManager")
    BiocManager::install("ipd")
    

  • Replaced CRAN build badge with:

    [![Bioc build status](https://bioconductor.org/shields/build/release/bioc/ipd.svg)](https://bioconductor.org/packages/ipd)
    

• New functions

  • ppi_a_ols(): implements the PPIa estimator for prediction‑powered inference.
  • chen_ols(): implements the Chen & Chen estimator for inference on predicted data.
  • .parse_inputs() - helper to validate and split input data.
  • .drop_unused_levels() - helper to drop unused factor levels and report which were removed.
  • .warn_differing_levels() - helper to warn on differing factor levels between labeled and unlabeled data.
  • .build_design() - helper to build design matrices and outcome vectors.
  • show() - implements S4 method for ipd class.

• Updates to functions

  • ipd() - registered new methods in wrapper.
  • ipd() - helper functions for parsing inputs and additional warnings to users when parsing formulas.
  • methods.R - Updated with new S4 class for ipd.

• Bioconductor submission prep

  • Branch created, tag v0.99.0 applied.
  • Will request CRAN archiving of the CRAN version upon successful Bioconductor acceptance.

ipd 0.1.4 (2025-01-07)

• Added a help topic for the package itself (man/ipd-package.Rd) via R/ipd-package.R and roxygen2

• Updated the documentation for ipd():

  • Provided a more explicit description of the model argument, which is meant to specify the downstream inferential model or parameter to be estimated.

  • Clarified that not all columns in data are used in prediction unless explicitly referenced in the formula argument or in the label argument if the data are passed as one stacked data frame.

• Updated the documentation for simdat() to include a more thorough explanation of how to simulate data with this function.

• simdat() now outputs a data.frame with a column named "set_label" instead of "set" to denote the labeled/unlabeled observation indicator.

ipd 0.1.3 (2024-12-03)

• Added a NEWS.md file to track changes to the package.

• Added a pkgdown site for the package.

• ipd() now allows for regression through the origin with intercept = FALSE argument.

• ipd() now takes an additional argument, na_action, to handle missing covariate data.

  • Currently supports "na.fail" and "na.omit". Defaults to na.fail.

  • Provides a more informative error message and lists which covariates are missing observations.

• ipd() now takes an additional argument, n_t, which denotes the (optional) size of the training set used to generate the prediction rule. Defaults to Inf but is necessary for the postpi_X methods if n_t < n, N, the number of labeled and unlabeled observations, respectively, in the data being analyzed.