bayesian_structural_time_series

This repo is my personal repo for learning how to use bayesian structural time series (bsts) for time series analysis. Using bsts as a foundation, one can perform causal inference of an intervention on time series data by modelling the past and using it as a counterfactual baseline.

Links to readings will be kept on this README document for reference.

Side Project!

1. Causal Analysis of PM 2.5 Air Quality in Los Angeles During Quarantine using CausalImpact and BSTS

Readings

Introductory:

• Sorry ARIMA but I'm going Bayesian
• Fitting Bayesian structural time series with the bsts R package
• Structural Time-Series Models
• Spike and slab: Bayesian linear regression with variable selection
• Making Causal Impact Analysis Easy

Package papers and tutorials:

• Predicting the Present with Bayesian Structural Time Series
• INFERRING CAUSAL IMPACT USING BAYESIAN STRUCTURAL TIME-SERIES MODELS
• R package bsts tutorial
• Causal Impact using Bayesian Structural Time-Series Models
• Software for Bayesian Structural Time Series
• MineThatData Forecasting Challenge: proposed solution with Bayesian Structural Time Series models

R packages

The following packages may be used in this learning repo:

• bsts
• CausalImpact
• MarketMatching
• tidyverse
• tidyquant

Datasets

For my analysis, I used the following data sources:

• Los Angeles Air Quality Data