Title
Jump Regression Analysis in R
Document Type
Student Presentation
Presentation Date
4-15-2019
College
College of Arts and Sciences
Department
Department of Mathematics
Faculty Sponsor
Dr. Partha Sarathi Mukherjee
Abstract
Regression analysis is a method of estimating the mean of response variable as a function of other explanatory variables. To use the simplest form, simple linear regression, one must make a strong assumption that the underlying function is linear. In literature, there are techniques such as local polynomial regression which only assumes that the underlying function is continuous. However, continuity assumption may still not be realistic in many situations. At first, I briefly mention a few methods to estimate the regression function when one can assume continuity, and then I describe two methods to estimate the regression function when there are finitely many jumps or discontinuities. In this project, I consider two cases: when the number of jumps is known, and when it is unknown. Next, I demonstrate how a practitioner can use these methods in various other situations, such as when one knows the prospective locations of the jumps, but not the total number of jumps. Finally, I show applications of some of these methods on a real data using my R functions.
Recommended Citation
Cox, Ryan, "Jump Regression Analysis in R" (2019). 2019 Undergraduate Research and Scholarship Conference. 39.
https://scholarworks.boisestate.edu/under_conf_2019/39