This advanced course delves into the world of Bayesian statistics, specifically focusing on time series analysis. It is designed for practicing and aspiring data scientists and statisticians who are familiar with calculus-based probability, maximum likelihood estimation, and Bayesian inference. Throughout the four-course sequence, students will build on their knowledge from previous courses to develop an understanding of modeling temporal dependencies, performing Bayesian inference and forecasting, and utilizing open-source software R for analysis and forecasting of time series.
The course covers various essential topics such as the AR(1) process, AR(p) process, normal dynamic linear models (NDLM), and their applications, along with utilizing R for simulations and examples. Students will learn about the principles of maximum likelihood estimation, Bayesian inference, and model order selection, as well as gain insights into the superposition principle, filtering, smoothing, and forecasting in NDLMs. Furthermore, the course includes a final project, allowing students to apply their newly acquired knowledge and skills in a practical setting.
With instructor Raquel Prado guiding the way, the course aims to equip learners with the necessary tools and techniques to effectively model temporally dependent data, implement specific classes of models, and perform Bayesian analysis for real-world datasets.
Certificate Available ✔
Get Started / More Info
This course consists of five modules that comprehensively cover various aspects of time series analysis, including the AR(1) process, AR(p) process, normal dynamic linear models, and a final project.
This module introduces students to the fundamental concepts of time series analysis, focusing on the AR(1) process. Students will gain insights into stationarity, autocorrelation function (ACF), partial autocorrelation function (PACF), maximum likelihood estimation, Bayesian inference, and the implementation of R for simulations and examples.
In this module, students will delve into the AR(p) process, exploring its definition, state-space representation, spectral representation, and Bayesian inference. The module also covers model order selection, ARIMA processes, properties of AR processes, and Bayesian analysis of an EEG dataset using an AR(p).
Week 3 focuses on normal dynamic linear models (NDLM) with a particular emphasis on polynomial trend models, regression models, filtering, smoothing, and forecasting. Students will gain practical experience using the dlm package in R and explore the sensitivity of NDLM to the model parameters.
This module continues the exploration of NDLM, covering Fourier representation, building NDLMs with multiple components, filtering, smoothing, and forecasting with unknown observational variance, and specifying the system covariance matrix via discount factors. Students will also engage in NDLM data analysis using real-world datasets.
The final module is dedicated to the culmination of the course with a final project. Students will have the opportunity to apply their knowledge and skills in a practical setting, showcasing their ability to build models, perform Bayesian inference, and utilize R for time series analysis and forecasting.
This course provides an introduction to computational statistics for data scientists, focusing on Bayesian inference using Python and PyMC3 for real-world scenarios....
Inferential Statistics introduces students to making inferences from sample relations to population, covering significance testing, statistical tests, and software...
Introduction to Statistical Learning covers concepts in statistical modeling, regression, classification, and more. Ideal for individuals with a broad range of undergraduate...
Model Diagnostics and Remedial Measures is a comprehensive course focusing on detecting and remedying violations of linear regression model assumptions. Participants...