Please check

https://docs.google.com/document/d/1Va_CnqUgMtGCAO6hRUENvu4F7j2KTsWAXSBxKP0OZ0M/edit?usp=sharing

for detail information. Below is a problem set that helps you decide whether you are ready for this course (draft version, don't write yet)

https://drive.google.com/file/d/1fWoYhHQmbVyyyupOJQp73sZDZRD_PZsK/view?usp=sharing

---

Econ 5166 is an introductory course to “classical” machine learning (ML) methods (e.g., decision trees, LASSO, …etc.) with an emphasis on their applications in social science research. This course is most suitable for students who have finished their first statistics course, have some experience in data manipulation, and would like to explore the ideas behind machine learning.

While there are already many great ML courses offered at NTU, our course distinguishes itself by two features. First, we will emphasize the link between machine learning and statistics. ML models, as well as the techniques (e.g., cross-validation), will be taught through the lens of statistics. Specifically, we will also cover hypothesis testing in a data-rich environment, a topic that is often ignored in typical ML courses.

Second, we will examine how and why ML is used in social science research. For each method covered in class, we will go over in-depth a research paper that uses that method to understand why there is a use case for ML. We will also write our own code (both in class and in homework assignments) to replicate and extend these researches, helping you gather experience in handling real data and sharpen your practical skills in data analysis.

This course will cover four ML topics: PCA and clustering, empirical Bayes and large-scale inference, LASSO and ridge regression, and tree-based methods. At the beginning of this course, we will also have a high-level introductory topic on quantitative research in social science that partly serves as a review of statistics and Python data analysis.