Extending Time-Series Models

PyMC3 is a probabilistic programming language that uses Bayesian statistics to specify models and estimate unknown quantities. PyMC3 has about seven time-series distributions, an important class of models in Probabilistic Programming. This project works on extending this class in a few ways: Firstly, the addition of a new model - namely Auto-Regressive Integrated Moving Average(ARIMA). Expansion of the existing distributions(and their respective documentation), including Time-Series analysis functions. Thirdly, state-space implementation of possibly a time-varying linear and Gaussian time series model, based on Aesara. Additionally, I believe that Programming can be an efficient way of understanding applied mathematics, motivating me to focus on more pedagogical or explanatory notebooks about Probabilistic Programming (Time-Series in this case). Explanatory Notebooks for ARIMA and state-space model usage, exploring their Econometric applications will be a small step towards fostering an interdisciplinary bond between academia and the open-source community, mutually benefiting both.