All Times EDT
The United States Census is used not only for apportioning the representation of legislators, but also for allocating resources for housing, highways, schools, assistance programs, and other projects that are vital to the prosperity, welfare, and economic growth. As such, it is important that we have a correct and accurate enumeration of those in the United States on census day. Turing to voter registration, it’s important that when voters go to polls, there is only one voter identification number associated with one voter to prevent against voter fraud. In order to get at the root of both applications, it’s important to understand that individuals are duplicated in data sources due to survey design and due to the way that our society collects more and more data. How does one remove such data in order to answer important questions facing society? What methods should be employed? What challenges does on face, and how do we overcome them. This talk will focus on two case-studies about the U.S. Census and voter registration, motivated by the data, where we present new methodology and algorithms, and speak to present solutions and future work in both of these areas.