Skip to content
Phase-amplitude coupling (PAC) toolbox
Python Other
  1. Python 99.2%
  2. Other 0.8%
Branch: master
Clone or download


Getting Started with pactools

Build Status Test coverage Python27 Python36

This package provides tools to estimate phase-amplitude coupling (PAC) in neural time series.

In particular, it implements the driven auto-regressive (DAR) models presented in the reference below [Dupre la Tour et al. 2017].

Read more in the documentation.


To install pactools, you first need to install its dependencies:

pip install numpy scipy matplotlib scikit-learn

To enable all features, you will also need to install optional packages:

pip install mne h5py

Then install pactools with one of the following two commands:

  • Development version:

    pip install git+
  • Latest stable version:

    pip install pactools

To upgrade, use the --upgrade flag provided by pip.

To check if everything worked fine, you can do:

python -c 'import pactools'

and it should not give any error messages.

Phase-amplitude coupling (PAC)

Among the different classes of cross-frequency couplings, phase-amplitude coupling (PAC) - i.e. high frequency activity time-locked to a specific phase of slow frequency oscillations - is by far the most acknowledged. PAC is typically represented with a comodulogram, which shows the strenght of the coupling over a grid of frequencies. Comodulograms can be computed in pactools with more than 10 different methods.

Driven auto-regressive (DAR) models

One of the method is based on driven auto-regressive (DAR) models. As this method models the entire spectrum simultaneously, it avoids the pitfalls related to incorrect filtering or the use of the Hilbert transform on wide-band signals. As the model is probabilistic, it also provides a score of the model goodness of fit via the likelihood, enabling easy and legitimate model selection and parameter comparison; this data-driven feature is unique to such model-based approach.

We recommend using DAR models to estimate PAC in neural time-series. More detail in [Dupre la Tour et al. 2017].


This work was supported by the ERC Starting Grant SLAB ERC-YStG-676943 to Alexandre Gramfort, the ERC Starting Grant MindTime ERC-YStG-263584 to Virginie van Wassenhove, the ANR-16-CE37-0004-04 AutoTime to Valerie Doyere and Virginie van Wassenhove, and the Paris-Saclay IDEX NoTime to Valerie Doyere, Alexandre Gramfort and Virginie van Wassenhove,

Cite this work

If you use this code in your project, please cite [Dupre la Tour et al. 2017]:

    author = {Dupr{\'e} la Tour, Tom and Tallot, Lucille and Grabot, Laetitia and Doy{\`e}re, Val{\'e}rie and van Wassenhove, Virginie and Grenier, Yves and Gramfort, Alexandre},
    journal = {PLOS Computational Biology},
    publisher = {Public Library of Science},
    title = {Non-linear auto-regressive models for cross-frequency coupling in neural time series},
    year = {2017},
    month = {12},
    volume = {13},
    url = {},
    pages = {1-32},
    number = {12},
    doi = {10.1371/journal.pcbi.1005893}
You can’t perform that action at this time.