Getting Started with reflowR

What is reflowR?

reflowR extends the workflowr package with themed colour schemes, structured analysis pipeline templates, and utility scripts for reproducible research. It wraps workflowr::wflow_start() and layers on additional structure and theming.

Installation

# Install from GitHub (workflowr is installed automatically)
if (!requireNamespace("pak", quietly = TRUE)) install.packages("pak")
pak::pak("CTTIR/reflowR")

# Required by the generated project setup
if (!requireNamespace("here", quietly = TRUE)) install.packages("here")

Quick Start

library(reflowR)

# Choose a new directory for the clinical 8-step project
dir.create("~/projects", recursive = TRUE, showWarnings = FALSE)
project <- reflow_init(
  directory = "~/projects/my_project",
  author    = "Your Name",
  email     = "[email protected]",
  scheme    = "clinical",
  depth     = "standard",
  change_wd = FALSE,
  open      = FALSE
)

Choosing a Scheme

reflowR schemes are the nine CTTIR v5 topic families:

The earlier names basic, code, special and other remain accepted as aliases of brand, methods, evidence and platform.

Preview any scheme:

reflow_preview("clinical")

Choosing a Depth

Project Structure

After reflow_init(), your project contains:

my_project/
+-- analysis/        # Rmd pipeline files (rendered to docs/)
|   +-- _site.yml    # Themed navbar + output config
|   +-- index.Rmd    # Home page with pipeline table
|   +-- about.Rmd    # Author card + pipeline docs
|   +-- 01_read_data.Rmd  # ... pipeline steps
|   +-- ...
+-- assets/          # Theme CSS, SCSS, and family icon
+-- code/            # R scripts
|   +-- 00_setup.R   # Packages, paths, colors, ggplot theme
|   +-- utils.R      # Helper functions
|   +-- formulary.R  # Reusable analysis recipes
|   +-- data_io.R    # Data import/export
|   +-- table_helpers.R  # Table builders
|   +-- plot_helpers.R   # Plot constructors
+-- data/raw/        # Raw data (never modify)
+-- data/processed/  # Cleaned data
+-- docs/            # Rendered HTML site
+-- output/          # Figures, tables, models
+-- report/          # Quarto report template

Working with Your Project

The returned project path makes the following commands work from the calling directory. Generated code/00_setup.R reports additional packages required by individual analysis helpers. Install them before using those helpers; they are not installed automatically. The optional table helpers use gtsummary >= 2.0.0 and require R >= 4.2, while core reflowR supports R >= 4.1. The initial scaffold supplies templates, not completed analyses.

After reviewing your analysis, use standard workflowr commands:

# Render all Rmd files
workflowr::wflow_build(project = project)

# Commit and publish
workflowr::wflow_publish(file.path(project, "analysis", "*.Rmd"),
                        message = "Initial build",
                        project = project)

# Check status
workflowr::wflow_status(project = project)

# View the site
workflowr::wflow_view(project = project)

# Set up GitHub Pages
workflowr::wflow_use_github("username", "repo", project = project)

Paired comparisons in the generated formulary

After sourcing code/formulary.R, wf_compare_groups() can compare two paired groups using an explicit participant or specimen identifier:

comparison <- wf_compare_groups(
  measurements, outcome = "value", group = "visit",
  paired = TRUE, pair_id = "participant_id"
)

Every ID must occur exactly once in each of the two groups, with a finite numeric outcome. Missing, duplicate or unmatched IDs are refused; row order never defines pairing. Factor levels define the first and second groups; otherwise group labels are sorted. Reported differences follow first minus second. IDs such as "001" and "1" remain distinct.

The existing automatic selector uses a Shapiro–Wilk check of the within-pair differences for 3–5,000 pairs. Otherwise it retains its t-test convention. The paired t test reports Cohen’s d standardized by the standard deviation of the differences; the Wilcoxon signed-rank test reports matched-pairs rank-biserial correlation. Both use the same ID-aligned observations. Constant or nonfinite differences are refused. This convenience selector does not validate study design, test assumptions or scientific inference; choose and justify the analysis before using its results. Unpaired calls retain their existing behavior. Existing generated projects must update code/formulary.R deliberately to receive this repair.

Customization

• CSS: Edit analysis/theme.css to customize styling
• New Rmd files: Add to analysis/ and update _site.yml navbar
• Navbar: Edit analysis/_site.yml
• ggplot theme: Modify theme_wf() in code/00_setup.R