Regrans

Segmented Linear Regression Models in R
Detect transition points by fitting two linear segments and selecting the best split.

---

Table of contents

• What is Regrans?
• Main features
• Installation
  • Install from GitHub
  • Install from package files
• Quick start
• How it works
• Functions overview
• Data included
• Documentation
• Authors
• License

---

What is Regrans?

Regrans is an R package for fitting segmented linear regressions (piecewise linear models with two segments).

It evaluates candidate split points along the explanatory variable, fits one linear model to the left side and another to the right side, and returns the split that minimizes the total residual sum of squares.

This is especially useful when your data suggests a transition point, breakpoint, or allometric inflection.

---

Main features

• ✅ Fit two-segment linear regressions from numeric vectors (x, y)
• ✅ Automatic search for the best split point
• ✅ Summary methods for left/right fitted models
• ✅ Built-in plotting helper for segmented fits
• ✅ ANCOVA helper to compare segment parameters
• ✅ Example dataset (simdata) included

---

Installation

Install from GitHub

# install.packages("devtools")
library(devtools)
install_github("rodrigosantana/Regrans")

Install from package files

Linux / macOS (.tar.gz)

Download:

• Regrans_0.0.1.tar.gz

Install from terminal:

R CMD INSTALL -l /path/to/your/R/library Regrans_0.0.1.tar.gz

Or from R:

install.packages("Regrans_0.0.1.tar.gz", repos = NULL,
                 lib.loc = "/path/to/your/R/library",
                 dependencies = TRUE)

-l and lib.loc are optional and only needed when installing into a custom user library.

Windows (.zip)

Download:

• Regrans_0.0.1.zip

Install from R:

install.packages("Regrans_0.0.1.zip", repos = NULL,
                 dependencies = TRUE)

---

Quick start

library(Regrans)

data(simdata)

# Fit segmented regression (minimum points in each segment = 5)
fit <- regrans(simdata$x, simdata$y, n.min = 5)

# Inspect left and right linear models
summary(fit)

# Plot points and fitted segment lines
plot_regrans(fit,
             col.lines = c("red", "blue"),
             lty.lines = c("solid", "solid"),
             lwd.lines = c(2, 2),
             pch = 16,
             xlab = "X",
             ylab = "Y",
             main = "Segmented linear regression")

# Optional: compare segment parameters via ANCOVA
ancova_fit <- regrans_ancova(fit)
summary(ancova_fit)

---

How it works

Regrans follows a simple and robust strategy:

1. Sort data by x (and y for ties).
2. Generate candidate split positions respecting n.min points per side.
3. For each split:
   • fit lm(Y ~ X) on the left subset,
   • fit lm(Y ~ X) on the right subset.
4. Compute residual sum of squares for both models.
5. Select the split with the smallest total residual sum of squares.

The returned object stores the two selected linear models (left and right), enabling plotting, summaries, and downstream inference.

---

Functions overview

Function	Purpose
regrans(x, y, n.min = 5)	Fits segmented linear regression and returns the best two-model split.
regranslm(x, y, n.min = 5)	Evaluates all candidate splits and returns complete fitting details.
summary.regrans(object, ...)	Summarizes left/right fitted linear models.
plot_regrans(obj, ...)	Plots observed data and both fitted segment lines.
regrans_ancova(obj)	Fits an ANCOVA model (Y ~ X * model) for segment comparison.

---

Data included

The package provides simdata, a simulated dataset with:

• 50 observations
• x: explanatory variable
• y: response variable

Use it to quickly test package workflows and examples.

---

Documentation

• PDF reference manual: Regrans-manual.pdf

---

Authors

• Rodrigo Sant'Ana
• Fernando Mayer
• Paulo Ricardo Pezzuto

---

License

Released under the GNU General Public License v3.0 (GPL-3).