|
Abstract:
|
Many resources are available for designing a first course in data science. What should be taught in a second course?
In this talk, we will describe the structure of a second course in data science. It was developed for a B.S. in Data Science program in which the use of existing coursework from other programs is maximized. We assume students have already been exposed to material in a general purpose statistics course, math through integral calculus, and programming in some language, as well as a first course in data science.
Data science is depicted as the intersection of hacking skills, statistics + math knowledge, and domain expertise in Drew Conway's famous Venn diagram. This course covers the first two with a Python-based practical overview of the purpose and assumptions of a wide variety of statistical methods including Bayesian, non-parametric and causal inference. Instead of focusing on applications from one particular domain, we give students the tools to combine domain knowledge with an appropriate statistical framework. Simulations are used throughout the course for computing, methodology evaluation, debugging, and building understanding.
|
