Metran
Metran is a package for performing multivariate timeseries analysis using a technique called dynamic factor modelling. It can be used to describe the variation among many variables in terms of a few underlying but unobserved variables called factors.
Installation
To install Metran, a working version of Python 3.7 or 3.8 has to be installed on your computer. We recommend using the Anaconda Distribution with Python 3.8 as it includes most of the python package dependencies and the Jupyter Notebook software to run the notebooks. However, you are free to install any Python distribution you want.
To install metran, type the following command
pip install metran
To install in development mode, clone the repository and type the following from the module root directory:
pip install -e .
Documentation
The docs can be viewed here.
Examples
For a brief introduction of the theory behind Metran on multivariate timeseries analysis with dynamic factor modeling see the notebook:
A practical real-world example, as published in Stromingen (Van Geer, 2015), is given in the following notebook:
A notebook on how to use Pastas models output with Metran:
Dependencies
Metran has the following dependencies which are automatically installed if not already available:
- numpy>=1.16.5
- pandas>=1.0
- scipy>=1.1
- matplotlib>=3.0
- pastas>=0.16.0
- numba
References
- Berendrecht, W.L. (2004). State space modeling of groundwater fluctuations.
- Berendrecht, W.L., F.C. van Geer (2016). A dynamic factor modeling framework for analyzing multiple groundwater head series simultaneously, Journal of Hydrology, 536, pp. 50-60, doi:http://dx.doi.org/10.1016/j.jhydrol.2016.02.028.
- Van Geer, F.C. en W.L. Berendrecht (2015) Meervoudige tijdreeksmodellen en de samenhang in stijghoogtereeksen. Stromingen 23 nummer 3, pp. 25-36.