GSOC 2024- Implement New Statespace Models

Hello! My name is Camilo Saldarriaga and I am finishing my master´s degree on Financial Economics at Paris 1- Panthéon Sorbonne.

I have knowledge and experience on time series models and I would like to contribute to this topic for the Google Summer of Code.

Hi Camilo!

I’m very glad your interested in participating in GSoC. Camilo was one of my Macroeconomics students last year, so I can vouch for his time-series chops.

For the statespace project, make sure you check out the example notebooks that already exist in the pymc-experimental library so you can see how these sorts of models work. The statsmodels documentation is also really good, and I copied them heavily when I developed the statespace module for PyMC. The developer who worked on that also wrote a paper to go with it, which is another good reference.

For more high level/academic topics, you can check out:

  1. This github repo on Kalman Filtering, which is the inner “guts” of how linear statespace models work
  2. This quantecon lecture, also on Kalman Filtering
  3. Time Series Analysis by Statespace Methods, by Durbin and Koopsman, is the textbook bible for these models. I basically had it open on a 2nd monitor at all times when working.

Your next step to getting involved will be to go onto our repositories – one of pytensor, pymc, or pymc-experimental – and making a PR. You can search for the tag “beginner friendly”, find an issue you think is interesting, and make a post saying you’d like to take it. From there, the devs will be available to help you through the process, so don’t be shy about asking questions.

1 Like

Thank you Jesse!! I will complete the PR during the next few days.

Thank you for your answer and for the welcome mesage.