snha4py
Snha4py is a Python implementation of the St. Nicholas House algorithm (Groth et al. 2019, Hermanussen et al. 2021, Hake et al. 2023). The algorithm infers a graph based on the correlation coefficient between variables of a data set. Here, the nodes and the edges would be the variables of the data set and the correlation coefficient, respectively. The correlation coefficient is evaluated in terms of association chains. Therfore, the absolute coefficient is ranked and compared pairwise. An association chain is characterized by a sequence of variables for which reversing the start and end point does not change the order of elements. These chains then build the graph.
Next, to the algorithm the package offers a variety of graph implementations, as well as a vizualization routine for the graphs and the correlation matrix.
Also, a module to create correlated data along a directed graph based on a Monte Carlo approach is included (Novine et al. 2022).
You can use the St. Nicolas House algorithm and other features of the package as a command-line interface (CLI) or you can import the package in your own code.
Documentation
Install
pip install snha4py
pip3 install git+https://github.com/thake93/snha4py.git --user
after the installation you should be able to run:
pip show snha4py
Create your 1st "St. Nicolas House"
Command-line interface
We start with creating a directory to store the test files.
mkdir test_snha
cd test_snha
In the first place we create test data in two steps:
- We create a directed graph.
python3 -m snha4py --create-graph -t werner -o werner.csv
- We create correlated data along the edges of the graph.
python3 -m snha4py --create-data --graph werner.csv -o data.csv --steps 25 --iterations 200 --plot
Now, we can run the algortihm to predict the graph based on the correlation of the data.
python3 -m snha4py --snha --data data.csv -o graph_pred.csv -p
Also, we are able to apply bootstrapping to increase the prediction (Hake et al. (2023)).
python3 -m snha4py --snha --data data.csv -o graph_pred_bt.csv --bootstrap -i 10 -p
Note: The result might differ from the picture below.
As import in your own code
python3
Python 3.x.xx ()
>>> from snha4py.Snha import Snha
>>> import matplotlib.pyplot as plt
>>> s = Snha()
>>> s.new_graph()
>>> s.create_corr_data(200,25)
>>> s.comp_corr(method='spearman')
>>> s.st_nich_alg()
>>> fig,ax = plt.subplots(1,3, figsize=(30,10))
>>> s.plot_graph(pred=False, ax=ax[0])
>>> s.plot_corr(ax=ax[1])
>>> s.plot_graph(ax=ax[2])
>>> ax[0].set_title('True Graph')
>>> ax[1].set_title('Correlation Coefficient')
>>> ax[2].set_title('Predicted Graph')
>>> plt.show()
References
Groth, D., Scheffler, C., and Hermanussen, M. (2019). Body height in stunted indonesian children depends directly on parental education and not via a nutrition mediated pathway - evidence from tracing association chains by st. nicolas house analysis. Anthropologischer Anzeiger, 76(5):445–451. DOI
Hake, T., Bodenberger, B., Groth, D. (2023). In Python available: St. Nicolas House Algorithm (SNHA) with bootstrap support for improved performance in dense networks. Human Biology and Public Health 4. (Accepted, doi follows)
Hermanussen, M., Aßmann, C., and Groth, D. (2021). Chain reversion for detecting associations in interacting variables—st. nicolas house analysis. International Journal of Environmental Research and Public Health, 18(4). DOI
Novine, M., Mattsson, C. C., & Groth, D. (2022). Network reconstruction based on synthetic data generated by a Monte Carlo approach. Human Biology and Public Health, 3. DOI