discord

discord is an R package that provides functions for discordant kinship modeling and other sibling-based quasi-experimental designs. It includes functions for data preparation, regression analysis, and simulation of genetically-informed data. The package is designed to facilitate the implementation of discordant sibling designs in research, allowing for the control of shared familial confounding factors.

Visit the discord website for more information and detailed documentation. Below is a brief overview of the package, its features, and a roadmap to get you started.

Quick Start Guide

Step 1: Install the Package

# Install from CRAN
install.packages('discord')

# Or install development version from GitHub
# install.packages('devtools')
devtools::install_github('R-Computing-Lab/discord')

Step 2: Choose Your Starting Point

Your workflow depends on your data structure and experience level:

🚀 New to discordant-kinship regression?

• Start with Full Data Workflow
• Demonstrates complete end-to-end example for beginners
• Transforms data from wide, long, or pedigree formats
• Selects siblings for OLS and orders for discordant analysis
• Shows all three models (OLS, Between-Family, Discordant) side-by-side
• Includes equations, manually specified syntax, as well as function calls

📊 Have NLSY data or existing kinship links?

• Use NLSY Regression Analysis
• Real-world example with flu vaccination and SES data
• Complete workflow from kinship linking to interpretation

🔧 Need to build kinship links from scratch?

• See Using discord with Simple Family Structures
• Construct links from basic family IDs (mother, father)
• Works without pre-existing kinship databases

Step 3: Explore Advanced Topics

Once you understand the basics, explore specialized topics:

• Categorical predictors: Categorical Predictors Vignette
• Visualizing results: Plotting Vignette
• Sample size planning: Power Analysis Vignette

Package Features

• Data Preparation: Functions to prepare and structure data for discordant sibling analysis, including handling of kinship pairs and demographic variables.
• Regression Analysis: Tools to perform discordant regression analyses, allowing for the examination of within-family effects while controlling for shared familial confounders.
• Simulation: Functions to simulate genetically-informed data, enabling researchers to test and validate their models.

Complete Vignette Roadmap

The package includes several vignettes organized by user needs. All vignettes can be accessed online or from the RStudio “Vignettes” tab after package installation.

📚 Start Here: Core Workflows

These vignettes provide complete end-to-end examples and should be your first stop:

• Full data workflow for discord
  • What you’ll learn:
    • How to transform data from wide, long, or pedigree formats
    • How to select siblings for standard OLS regression
    • How to discord orders siblings for discordant-kinship analysis
    • How to run and compare all three model types (OLS, Between-Family, Discordant)
      • including specify models using equations, manual syntax, and function calls
    • How to interpret difference scores and mean scores
    • Complete side-by-side model comparisons
• NLSY regression analysis with discord
  • Use this vignette if you want an end-to-end applied example that links NLSY79 relatives, cleans variables for flu vaccination and SES, constructs dyads, and then fits within-family models.
  • You will learn how to specify discord_regression correctly and interpret coefficients.

🔧 Data Preparation

• No Database? No Problem: Using discord with simple family Structures
  • This vignette is particularly useful for situations when you do not have existing kinship links and need to build relationships directly from simple family identifiers.
  • It shows how to construct the links, optionally simulate phenotypes under specified structures, and fit discord_regression with alternative specifications for small or bespoke datasets.

📊 Advanced Topics

• Creating plots for discord
  • This vignette takes fitted discord_regression outputs and produces publication-ready ggplot figures of effect estimates and within-family contrasts with minimal transformation of the model results.
  • It includes complete plotting code paths you can reuse, from extracting estimates to saving figures that clearly communicate within-family findings.
• Power Analysis with discord
  • Use this vignette when you need to plan sample sizes or evaluate power by running simulation grids that vary effect sizes, kin types, and Ns using kinsim, then re-fitting discord_regression under each condition. It reports empirical power and writes tidy summaries.
• Handling categorical predictors with discord
  • This vignette formalizes categorical predictors in discord designs by separating categorical variables into within-dyad and between-dyad components. It makes the implied contrasts explicit.
    
  • It discusses the pitfalls of interpreting coefficients when using categorical predictors, and reviews best practices for coding and interpretation.

External Reproducible Examples

Beyond the vignettes, you can find additional examples that fully reproduce analyses from our other publications (Garrison et al 2025, etc). These examples can be accessed via the following links and are presented in reverse chronological order:

• National Longitudinal Survey of Youth (NLSY) datasets

  • NLSY AMPPS repo: Reproduces NLSY analyses from Garrison et al 2025, using targets for workflow management. Garrison, S. M., Trattner, J. D., Lyu, X., Prillaman, H. R., McKinzie, L., Thompson, S. H. E., & Rodgers, J. L. (2025). Sibling Models Can Test Causal Claims without Experiments: Applications for Psychology. https://doi.org/10.1101/2025.08.25.25334395

  • Frontiers repo: Reproduces Sims, E. E., Trattner, J. D., & Garrison, S. M. (2024). Exploring the relationship between depression and delinquency: a sibling comparison design using the NLSY. Frontiers in psychology, 15, 1430978. https://doi.org/10.3389/fpsyg.2024.1430978

  • Intelligence repo: Reproduces Garrison, S. M., & Rodgers, J. L. (2016). Casting doubt on the causal link between intelligence and age at first intercourse: A cross-generational sibling comparison design using the NLSY. Intelligence, 59, 139-156. https://doi.org/10.1016/j.intell.2016.08.008

• China Family Panel Studies (CFPS) dataset
  • CFPS AMPPS repo: Reproduces analyses from the China Family Panel Studies (CFPS) dataset, focusing on the association between adolescent depression and math achievement. Garrison, S. M., Trattner, J. D., Lyu, X., Prillaman, H. R., McKinzie, L., Thompson, S. H. E., & Rodgers, J. L. (2025). Sibling Models Can Test Causal Claims without Experiments: Applications for Psychology. https://doi.org/10.1101/2025.08.25.25334395

Citation

If you use discord in your research or wish to refer to it, please cite the following package as well as the AMPPS paper:

To cite package 'discord' in publications use:

  Garrison S, Trattner J, Hwang Y (2026). _discord: Functions for
  Discordant Kinship Modeling_. doi:10.32614/CRAN.package.discord
  <https://doi.org/10.32614/CRAN.package.discord>, R package version
  1.3, <https://github.com/R-Computing-Lab/discord>.

A BibTeX entry for LaTeX users is

  @Manual{,
    title = {discord: Functions for Discordant Kinship Modeling},
    author = {S. Mason Garrison and Jonathan Trattner and Yoo Ri Hwang},
    note = {R package version 1.3},
    url = {https://github.com/R-Computing-Lab/discord},
    year = {2026},
    doi = {10.32614/CRAN.package.discord},
  }

Contributing

Contributions to the discord project are welcome. For guidelines on how to contribute, please refer to the Contributing Guidelines. Issues and pull requests should be submitted on the GitHub repository. For support, please use the GitHub issues page.

License

discord is licensed under the GNU General Public License v3.0. For more details, see the LICENSE file.