robarma 0.1.0

Python bindings for the robarma C++ library

RobARMA Python

Python bindings for the RobARMA C++ library.

About this repository

This repository provides a modern, pip-installable Python interface to the RobARMA C++ library, which implements robust and classical estimators ARMA(p, q) processes.

Features:

• Robust estimators:
  • S
  • FTAU (filtered tau)
  • MM
  • BIP-MM (bounded innovation propagation MM)
• Classic estimators:
  • OLS (ordinary least squares)
  • MLE (maximum likelihood via Kalman filter)

Installation

Clone this repository and initialize submodules:

git clone https://github.com/kustij/robarma-py.git
cd robarma-py
git submodule update --init --recursive

Building

Ceres and Eigen are required. Simplest way is to use vcpkg and pass it to pip install as

pip install . --config-settings=cmake.args="-DCMAKE_TOOLCHAIN_FILE=/path/to/vcpkg/scripts/buildsystems/vcpkg.cmake"

Usage

import robarma
import numpy as np

# Simulate an ARMA(1, 1) process
y = robarma.simulate(phi=np.array([0.5]), theta=np.array([0.2]), mu=1.0, n=100)

# Fit an ARMA(1, 1) model using OLS
model = robarma.arma_model(y, 1, 1)
fit = robarma.mm(model)
print(fit.params)

Documentation

• Python docstrings are available for all major classes and functions.
• For full C++ API documentation, see the robarma Doxygen docs.

License

See LICENSE file. The underlying C++ library is licensed under the same terms as robarma.