dokato / connectivipy

python module for connectivity analysis

ConnectiviPy

Python module for connectivity analysis. You may find here implementation of MVAR based estimators like DTF, PDC and also Coherency, PSI. For more information visit documentation page.

The project was supported by Google Summer of Code 2015 and done in collaboration with International Neuroinformatics Coordination Facility.

Content

• Data - data object that makes data handling easier
• Connectivity - class with connectivity estimators
• Mvar fitting - this submodule includes some MVAR algorithms

License

BSD 2-clause

Documentation

Visit ReadTheDocs for detailed documentation and tutorials.

Installation

Option 1: PIP (stable release)

pip install connectivipy

Option 1: PIP+GIT (latest release)

pip install git+https://github.com/dokato/connectivipy.git

Option 3: using GIT (the most recent version)

git clone https://github.com/dokato/connectivipy.git
cd connectivipy
python setup.py install

Authors

• Dominik Krzemiński
• Maciej Kamiński (scientific lead)

connectivipy's Issues

Labels on plot conn

it should have better allocation of axes labels and x,-y-ticks.

`connectivipy.aec` Module not found, install from source

Fixed by adding an __init__.py to aec submodule and adding connectivipy.aec to packages list in setup.py.

[feature] edf loading

euroepean data format

New estimators

Recently some new approaches appeared. It'd be worth to review and try to implement them.

add colorbar to short-time subplots

Just like here: http://stackoverflow.com/questions/13784201/matplotlib-2-subplots-1-colorbar

bug in class mvarmodel.Mvar

in doc should be fitting_algorithm not:

possible methods are in fit_dict where key is acronym

docstrings in mvar_gen and mvar_gen_inst are wrong

      *y* : numpy.array
          (k, n-omit) data points

should be n
2)

*Acf* : numpy.array
          array of dimension kxkxp 

should be shape (p,k,k)

example with trans3d

to show users how to make 3d maps with arrows

assertion in short time

assert nfft>no, "overlap must be smaller than window"

some minor things

• license stuff
• not alphabetical order in sphinx

autodoc_member_order = 'bysource'

__calc_multitrial uses wrong data dimension size

__calc_multitrial from ConnectAR defines the number of trials as:
trials = data.shape[0]
whereas the documentation states that the trial number is the third dimension, i.e.
trials = data.shape[2]
hence, it leads to an error if the no of channels is > than the no of trials.

write more nose tests

add spectral G-causality

From http://www.scholarpedia.org/article/Granger_causality

percent of maximum in short-time plotting

to improve clarity of the time-frequency map

[ENH] Correlation of envelopes

New method.

Python 3.11 ?

I've enjoyed using this package for a few years now, but I'm unable to install it on a Windows machine running Python 3.11. I know you've tested it with much earlier versions of Python, but I had it working fine on 3.9.9. Any chance I can easily correct the error below?

Typing pip3 install connectivipy returns a fairly long error,

  **Getting requirements to build wheel ... error
  error: subprocess-exited-with-error
  × Getting requirements to build wheel did not run successfully.**

and ending with

   **File "C:\Users\fishbacp\AppData\Local\Temp\pip-install-c7i1nbit\connectivipy_18c81df64274429a9b41c91627ce597c\connectivipy\data.py", line 7, in <module>
      import numpy as np
  ModuleNotFoundError: No module named 'numpy'**

I've installed numpy so that's not the issue.

values on edges

in plot values should be visible only on edges

How to cite?

I'm incorporating connectivipy into some software I'm writing that will look at network connectivity measures and be appropriate for use in a clinical setting. All python is converted to .app and .exe using pyinstaller. The software uses MNE, tkinter, and pandas extensively.

The package will be free of course, and my academic institution will help me construct an appropriate legal disclaimer. However, since I'm incorporating connectivipy, particular for MVAR construction, I'd like to know the appropriate citation you'd like me to use.

[bug] plotting style issues in python 3

It worked well in 2.7 but in 3.3 is unreadable for many channels.

[feature] stability of mvar

before mvar fitting algorithm should check stability of the paramters matrix A

remove unnecessary imports from init

main init in connectivipy main folder

short-time statistic

now it calculates only for one trial case

update documentation

new changes have been made and docstrings are not up to date

travis-ci not passing

Maybe some conda version changed (worth to try different mirror)

test trans3d export

because it's not working very well now

docstrings of statistics

surrogate/bootstrap and wrappers

plott_conn move to main

make new file named main.py and there put other module functions

bug in statistic

should be (1-alpha)*100 in levels

then it can be:

sc[0] = (1-alpha)*100
sc[1] = alpha*100

for twosided case

Significance levels

Significance levels testing the hypothesis of non nullity (e.g. PDC(freq) > 0) should be done over the probability distribution of the variable of interest centered in its estimate, i.e.

instead of

where P_{1-alpha} is the 1-alpha percentile of the resampled values, theta hat is the variable estimator and theta hat star are the resampled estimator values.
Ideally, this should be done pivoting by the bootstrapped sample standard deviation, i.e., testing the distribution of:

Although this is not always computationally feasible.
Also, the code
ficance[i][j] = np.max(st.scoreatpercentile(signi[:, :, i, j], (1-alpha)*100, axis=1))
in conn.Connect.levels, calculates the percentile across a trial's spectral values and takes the max across trials, instead of the converse (which would be achieved by axis=0).
Referece: Zoubir, Abdelhak M., and D. Robert Iskander. Bootstrap techniques for signal processing. Cambridge University Press, 2004.
Btw, I wish you all safety during this pandemic, It's been hard doing research during these difficult times.

add covariance regularize (Tikhonov regularization)

The nan value appears in order criterion values. Tikhonov regularization may solve this problem.
best, crit = mv.order_akaike(data, 15, 'vm')

remove pdb

there are e.g. in conn.py

python3 support

[bug] problems with pip installation in python3

Such error occurred:

      File "/usr/local/lib/python3.4/dist-packages/connectivipy/working.py", line 145
        print abs(gcval[:,1,0][0])+abs(gcval[:,2,0][0])+abs(gcval[:,2,1][0])
                ^
    SyntaxError: invalid syntax
   

to be checked why...

[bug] plotting python 3

check image formatting in python 3 - maybe some changes in matplotlib versions occured

Understanding the outcomes of Google Summer of Code

Dear Dominik Krzemiński,

I am a Ph.D. student at the University of São Paulo (Brazil). We would like to understand how and how much Open Source Communities benefit from participating in Google Summer of Code (GSoC). We believe that your experience in GSoC is valuable and could help us achieve that. In a 1-minute questionnaire, we would like to know:

Concerning the project entitled "ConnectiviPy - python module for brain connectivity analysis based on MVAR" developed in GSoC 2015.

1. Do you have a GitHub repository for the GSoC project? If so, could you provide us with the link to the project?

2. Have you continued committing code or opening issues to the GSoC project after GSoC? (Yes / No) If so, code, issues or both?

3. Have you continued committing code or opening issues to other OSS projects after GSoC? (Yes / No) If so, code, issues or both?

4. Could we use aggregate data from your participation in GSoC and other OSS projects in our research?

The following link contains a sample of our work and how we would use your data:
https://ieeexplore.ieee.org/document/8094410/

Regards,

Jefferson