Estimating time series models by state space methods in Python: Statsmodels

This repository houses the source and Python scripts producing the paper "Estimating time series models by state space methods in Python: Statsmodels".

Paper PDF

A PDF version of the paper can be found in the repository, and also at: https://github.com/ChadFulton/fulton_statsmodels_2017/raw/master/fulton_statsmodels_2017_v1.pdf

Notebooks

There are three Jupyter notebooks with code showing maximum likelihood and Bayesian estimation of three example models:

• ARMA(1, 1) model
• Local level model
• Simple Real Business Cycle model

Build

The paper is written using Sphinx. In particular, see:

• paper/source for the reStructuredText files of text
• paper/source/sections/code for all of the code that is referenced in the text and that produces the output and figures. To run all code and produce all output, run python run_all.py in that directory.
• notebooks for Jupyter notebooks that flesh out the three examples in the paper (ARMA(1, 1), local level, and a simple real business cycle model)

To build the paper, in a terminal from the base directory, you must:

>>> cd paper/source/sections/code
>>> python run_all.py
>>> cd ../../../
>>> make html
>>> make latex