This course has two broad objectives. One is to give you more experience in analysis of data. The other is to add to your took kit of statistical methods. The course emphasizes learning by doing. Applications in the course demonstrate that analysis of data can be exciting (really!). The course is also designed to help you understand and tackle the issue of causality. For some prediction and decision making problems, causality is not particularly important. For example, if I can accurately predict which movies individuals will like based on their past viewing choices, I may not care what causes them to pick the movies they do. For other decision problems, causality matters, confronting us with the familiar axiom that correlation does not imply causation. Suppose a new sales manager observes that sales and advertising have increased hand‐in‐hand in past years. Can he conclude that advertising caused sales to increase, or did increased sales lead to allocation of more money for advertising? A new mayor of a city observes that cities with high crime rates tend to have higher police expenditures per capita than cities with low crime rates. Can she get behind this correlation to assess how crime in her city will be affected if she increases the police force in her city? The course will cover methods that have been developed to disentangle causal effects, and the course will help you understand conditions under which those methods “work”. In addition to studying and evaluating applications done by others, students will be asked to work in groups to choose a question, gather relevant data, and do the analysis required to answer the question.
Lecture: 100min/wk and Recitation: 50min/wk