pymgcv 0.0.1

pymgcv: Generalized additive models using mgcv, in Python.

pymgcv: Generalized Additive Models in Python

pymgcv provides a Pythonic interface to R's powerful mgcv library for fitting Generalized Additive Models (GAMs). It combines the flexibility and statistical rigor of mgcv with the convenience of Python's data science ecosystem.

Currently in development. As this is a multilanguage project (R and Python), we use pixi, a package management tool which supports this (via conda). For development, the pymgcv can be installed by installing pixi and running:

git clone https://github.com/danielward27/pymgcv.git
cd pymgcv
pixi shell --environment=dev

Installation options

Installing the python package only includes the python package and dependencies. This means an R installation with mgcv is also required. Conda and pixi provide two convenient options for handling both Python and R dependencies:

Using conda:

• Install conda e.g. miniforge
• Install uv (or use pip).

conda create --name my_env python r-base r-mgcv
conda activate my_env
uv pip install pymgcv

Using pixi:

• Install pixi

pixi init
pixi add python r-base r-mgcv
pixi add --pypi pymgcv
pixi shell

Using either method the below example should now run e.g. in the terminal after running python, or in an IDE after selecting the pixi/conda environment.

Simple example

import pandas as pd
import numpy as np
from pymgcv.gam import GAM
from pymgcv.terms import S, T, L
from pymgcv.plot import plot_gam
import matplotlib.pyplot as plt

# Generate sample data with non-linear relationship
np.random.seed(42)
n = 100
x0 = np.random.uniform(-1, 1, n)
x1 = np.random.uniform(-1, 1, n)
y = 0.5 * x0 + np.sin(np.pi * x1) + np.random.normal(0, 0.5, n)
data = pd.DataFrame({'x0': x0, 'x1': x1, 'y': y})

# Define model: linear effect of x0, smooth function of x1
model = GAM({'y': L('x0') + S('x1')})

# Fit the model
model = model.fit(data)
plot_gam(fit=model, residuals=True)
plt.show()