Jasjeet Sekhon, Professor

Closed (1) Donor Networks in American Politics

Applications for Fall 2017 are now closed for this project.

This project will involve two components. First, using publicly-available data, URAP participants will construct a map or network of the flows of money between donors, foundations, political organizations, and campaigns in American politics. Second, URAP participants will help conduct surveys of elected officials, donors, and political organizations.

Helping to conduct a survey of elected officials, donors, and political organizations; gathering data to construct a network of money flows; data analysis if students have such skills; reviewing the literature for previous research; writing reports.

Regular meetings about the project will take place between students, Professor Sekhon (political science and statistics), and Josh Kalla (PhD candidate in political science). These meetings will give students an opportunity to learn about research design and data science (from Sekhon and Kalla).

Day-to-day supervisor for this project: Josh Kalla, Graduate Student

Qualifications: n/a

Weekly Hours: 3-6 hrs

Related website: http://sekhon.berkeley.edu

Closed (2) Developing Machine Learning Algorithms For Causal Inference In Big Data

Applications for Fall 2017 are now closed for this project.

With the rise of big data, there is growing interest in targeting treatments, programs, and policies to the individuals they would help the most and harm the least. Statistical algorithms are already being used to assign people to online ads, credit, and even prison. But also in medicine, physicians try to determine the individual treatment effect of a single patient to choose the best treatment.

In this project, we will develop and implement a new algorithm to estimate the treatment effect of a single unit (e.g. a patient). We will start with an idea based on the machine learning algorithm, Random Forest. The goal is to implement and publish a version of our algorithm suited for big data.

Regular meetings about the project will take place between students, Professor Sekhon (political science and statistics), and Sören Künzel (Ph.D. candidate in statistics). These meetings will give students an opportunity to learn about causal inference, machine learning, statistical inference for big data, and parallel computing.

Day-to-day supervisor for this project: Sören Künzel, Ph.D. candidate

Qualifications: Implementation will happen in C/C++ with bindings to R. Familiarity with C/C++, and basic knowledge of R is required. An understanding of machine learning techniques, such as Random Forest, Gradient Boosting, and Neural Networks, is appreciated but not required.

Weekly Hours: 9-12 hrs

Related website: http://sekhon.berkeley.edu