rgram 0.1.1

A Python library for regressograms and kernel smoothing, designed for data analysis and visualisation.

Rgram

Rgram is a Python library for performing regression analysis and visualisation. It provides tools for creating regressograms (rgrams) and performing kernel smoothing using Polars, a high-performance DataFrame library. The library is designed to simplify data analysis workflows and is compatible with uv for dependency management.

Notes on regressograms can be found in section 4.4 of García-Portugués, E. (2023). Notes for nonparametric statistics. Carlos III University of Madrid: Madrid, Spain.

Features

• Regressogram (rgram): Analyse relationships between variables with support for binning by index or distribution and optional Ordinary Least Squares (OLS) regression calculations.
• Kernel Smoothing (kernel_smoothing): Perform kernel smoothing using the Epanechnikov kernel for regression analysis.
• Flexible API: Designed for ease of use and high performance thanks to Polars DataFrames and LazyFrames.

Requirements

• Python >= 3.11
• uv for dependency management

Installation

To get started with Rgram, follow these steps:

1. Clone the repository:

   git clone https://github.com/JackGreenaway/Rgram.git
   cd Rgram
   

2. Install dependencies using uv:

   uv install
   

3. Verify the installation:

   uv sync
   

Usage

---

Example: Regressograms and Kernel Smoothing with Polars

Imports and setup

import polars as pl
import numpy as np
import matplotlib.pyplot as plt

from rgram.rgram import Regressogram, KernelSmoother

plt.style.use("ggplot")

Generate sample data

n = 50
x = np.sort(np.random.normal(0, 1, n))
y = 1 + x
y_noise = y + np.random.normal(0, 2, n)

Fit regressogram to noisy data

df = pl.DataFrame({"x": x, "y": y, "y_noise": y_noise})

rgramer = Regressogram(data=df, x="x", y="y_noise")
rgram = rgramer.fit_transform().collect()

fig, ax = plt.subplots(figsize=(6, 5))

ax.plot(x, y, lw=0.5, label="true function")
ax.scatter(x, y_noise, s=15, alpha=0.3, marker="o", color="black")
ax.step(rgram["x_val"], rgram["y_pred_rgram"], lw=0.5, label="rgram")
ax.fill_between(
    rgram["x_val"],
    rgram["y_pred_rgram_uci"],
    rgram["y_pred_rgram_lci"],
    alpha=0.2,
    label="ci",
)

ax.set_xlabel("x variable"), ax.set_ylabel("y variable")
ax.legend()
fig.tight_layout()
plt.show()

Kernel smoothing on regressogram output

smoother = KernelSmoother(data=rgram, x="x_val", y="y_pred_rgram")
ks_rgram = smoother.fit_transform().collect()

fig, ax = plt.subplots(figsize=(6, 5))

ax.plot(x, y, lw=0.5, label="true function")
ax.scatter(x, y_noise, s=15, alpha=0.3, marker="o", color="black")
ax.plot(ks_rgram["x_eval"], ks_rgram["y_kernel"], lw=0.5, label="smoothed rgram")

cis = []
for col in ["y_pred_rgram_lci", "y_pred_rgram_uci"]:
    smoother = KernelSmoother(data=rgram, x="x_val", y=col)
    cis += [smoother.fit_transform().collect()]

ax.fill_between(
    cis[0]["x_eval"],
    cis[0]["y_kernel"],
    cis[1]["y_kernel"],
    alpha=0.2,
    label="smoothed ci",
)

ax.set_xlabel("x variable"), ax.set_ylabel("y variable")
ax.legend()
fig.tight_layout()
plt.show()

This example demonstrates how to use the Regressogram and KernelSmoother classes to create a regressogram and apply kernel smoothing for visualisation.
Both classes follow a scikit-learn-like API with fit(), transform(), and fit_transform() methods.
For most use cases, fit_transform() is the recommended entry point.

---