I am currently working on running variance ratio tests in Python. Are you aware of any packages that are equivalent to Matlab's vratiotest (https://www.mathworks.com/help/econ/vratiotest.html) or R's vrtest (https://cran.r-project.org/web/packages/vrtest/index.html)?
I found several discussions on stackoverflow, however, none of these were helpful. The only package I am aware of is the arch package (https://github.com/bashtage/arch/blob/main/examples/unitroot_examples.ipynb). Any further ideas?
In the computational milieu for econometric and financial time-series analyses, variance ratio tests are fundamental in ascertaining the stochastic properties of a financial series, including its weak-form market efficiency. Although MATLAB's vratiotest and R's vrtest packages have cemented themselves as staples for performing such tests, Python's ecosystem has, heretofore, demonstrated a paucity of equivalent toolsets. However, the nascent yet burgeoning state of Python's econometric libraries proffers some avenues for undertaking these analyses.
The arch package, as you've noted, does extend some functionalities that are tangential to variance ratio tests, especially within its unit root testing suite. Its primary focus, though, is conditional heteroskedasticity models rather than time-series randomness or market efficiency per se.
For a more bespoke solution, one might consider leveraging Python's scientific stack—predominantly NumPy for numerical operations and SciPy for statistical methods—to implement the variance ratio test from first principles. This offers the dual advantage of methodological transparency and algorithmic extensibility, albeit at the expense of computational efficiency and out-of-the-box functionality.
Another potential avenue worth exploration is the statsmodels library, which, although not offering a variance ratio test explicitly, does provide a comprehensive suite of econometric tools. One could potentially appropriate its existing methods to construct a customized variance ratio testing paradigm. Given the extensible architecture of statsmodels, this would integrate seamlessly with other econometric and statistical procedures in Python.
Furthermore, emerging Python libraries for econometrics and quantitative finance may eventually incorporate variance ratio tests or analogous methodologies. These libraries, often found on repositories like GitHub, are generally in the alpha or beta stage but may offer preliminary or experimental implementations of the variance ratio test.
In summation, while Python currently lacks a dedicated, one-to-one mapping to Matlab's vratiotest or R's vrtest, its expansive and modular scientific ecosystem provides ample tools for either rolling out custom implementations or leveraging existing but not fully analogous functionalities. This current lacuna in Python's package repository certainly underscores an opportunity for future contributions to its quantitative finance and econometric libraries.
Hey Volker Seiler have a look at this implementation: https://github.com/hiemal/vratiotest/blob/master/vratiotest.py
Hi, many thanks for your answers. As indicated, I am aware of the arch package. Likewise, I had a look at the implementations provided by Mingze Gao, Lautaro Parada and the one indicated by Christian Schmidt. I'll go with the arch package. Again, many thanks!
- Badges
- Science topic
- Similar topics
- Statistics
- Survey