Linear Regression Library

A robust and lightweight C# library for performing 2D (bivariate) linear regression analysis. This project calculates statistical properties such as slope, y-intercept, correlation coefficient, and p-value for paired x, y data. It is designed for technical users, including data scientists, statisticians, and developers working on statistical modeling.

Overview

The library provides a Regression class for computing 2D regression statistics and a PValueStat class for calculating the p-value of the regression slope. It supports adding, removing, and merging data points, with robust handling of edge cases like insufficient data or NaN values.

Lineage

• Originated from 1978 TI-58 and TI-59 calculator algorithms.
• Ported to C in 1982, then to Perl 3 in 1987.
• Modernized in C# with updates in 2001, 2002, 2003, 2008, 2012, 2016, 2018, 2019, 2022, 2025
• Copyright © Jillian England, 2001–2022.

Features

• Data Management:
  • Add individual (x, y) data points or merge multiple regression datasets.
  • Remove data points (decrement) with support for both explicit y-values and implicit y (based on sample count).
  • Track minimum and maximum x and y values.
• Statistical Calculations:
  • Sum of x, y, x², y², and x*y.
  • Mean of x and y.
  • Standard deviation and variance for x and y (with N or N-1 weighting for population or sample studies).
  • Slope, y-intercept, and correlation coefficient.
  • P-value for the regression slope (via PValueStat class).
• Robustness:
  • Handles edge cases (e.g., insufficient samples, division by zero, NaN, or infinite values).
  • Returns double.NaN for invalid computations instead of throwing exceptions in most cases.
• Serialization:
  • The Regression class is marked [Serializable] for persistence.

Installation

The library is a C# project that can be included in your .NET solution:

1. Clone or download the repository.
2. Add the project or source files (Regression.cs, PValue.cs) to your solution.
3. Ensure the MathNet.Numerics NuGet package is installed for PValueStat (for t-distribution calculations).

dotnet add package MathNet.Numerics

License

This project is licensed under the Affero General Public License (AGPL-3.0). See the LICENSE file for full details.

Contact

Jillian England

• Email: jill.england@comcast.net
• X: @Firefox_XB9R
• FB: Jillian England Seattle
• Any omissions or errors in this software specification are unintentional and should be reported for correction.