- Tutorials written with a focus on atmospheric science applications as part of the Data Analysis Tools for Atmospheric Scientists (DATAS) Gateway. DATAS - https://datasgateway.colostate.edu/
- The methods explained in this repository are focused on observational studies where controlled experiments (e.g., targetted modelling studies in climate) are not performed to identify causes and effects. These methods allow you to identify 'potential' relationships that need to be further validated with our existing knowledge of a specific application domain.
- Created by Savini M. Samarasinghe, Colorado State University, Fort Collins, CO.
This is the most commonly used approach to find cause-effects in climate science to date.
- Built using Python 3. Requirements: numpy, pandas, matplotlib, scipy, statsmodels.
- Cite as: Savini Samarasinghe; Marie McGraw (2019), "Bivariate Granger Causality Example," https://datasgateway.colostate.edu/resources/234.
PC stable algorithm can be used to learn a probabilistic graphical model representation of data where the variables of interest are presented as nodes of a graph and the stochastic relationships between the variables are presented as graph edges.
About the files and requirements:
PC_stable_for_time_series.ipynbis the main tutorial. This notebook provides a simple example of how the PC stable algorithm can be used to find potential cause-effect relationships between a set of time series variables.Seasonal_data_extraction.ipynbgives an example of how to extract seasonal data. This notebook uses data fromsample_data.mat- Built using Python 3. Requirements: numpy, pandas, matplotlib, scipy, graphviz.
- Graphviz installation instructions: https://pypi.org/project/graphviz/
- Cite as: Savini Samarasinghe (2019), "PC Stable Example," https://datasgateway.colostate.edu/resources/218.