CCM Analysis for Python Notebooks

A Python package for performing Convergent Cross Mapping (CCM) analysis on time series data.

Note: This package is optimized for Jupyter Notebooks (.ipynb). If you need to perform CCM analysis in a Python script (.py), please use CCM_analysis_Python.

Installation

You can install the package via pip:

pip install ccm_analysis_pynb

Usage

To use this package in a Jupyter Notebook, import the function and run the analysis:

from ccm_analysis_pynb import run_ccm_analysis_jupyter

df = pd.DataFrame(data)

# Run CCM analysis
result_holder = run_ccm_analysis_jupyter(df, L=100, E=2, tau=1, THRESHOLD=0.8, save_output=False)

# Access the final DataFrame of significant relationships
final_results = result_holder.result
print(final_results)

Parameters

• data_input: Either a filepath (str) to tab-separated data or a pandas DataFrame.
• L: Length of time series to consider.
• E: Embedding dimension.
• tau: Time delay.
• THRESHOLD: Significance threshold for cross-map scores.
• save_output: Whether to save plots and protocol (default: True).
• output_dir: Directory to save outputs (default: current working directory).

Output

The function returns an interactive widget for manual evaluation of relationships. Once the evaluation is completed, it generates a final heatmap and saves the results if save_output=True.

Test Dataset

You can test the ccm_analysis package using the sample dataset provided in this repository. You can download the test dataset from the following link:

• Test Dataset (ccm_test_data.txt)

Data Format

The dataset is structured as follows:

• Columns represent the variables you want to analyze.
• Rows represent the time points for each variable.

Dependencies and Acknowledgements

Parts of this project are based on the Convergent Cross Mapping (CCM) implementation from the following repository from Prince Javier :

• Convergent Cross Mapping GitHub Repository Prince Javier

I have utilized parts of the CCM algorithm from this repository to help analyze causality in time series data in my own project.