Rounding error in least squares approximation with applications in financial mathematics
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Currently, there are many numerical techniques which can be used to estimate Least-Squares polynomial approximations to model interest rates. Four of these techniques are the Least-Squares approximation, the QR Decomposition approach to solving the Least-Squares problem, the Gram-Schmidt Orthogonalization Process approach to solving the Least-Squares problem and the Discrete Legendre Polynomial approach to solving the Least-Squares problem. Each of these four approaches is studied herein. A comparative study is used throughout to determine which has the least rounding error. Chapter II breaks down each of these four methods into their many parts and explains them with a mathematical approach. This chapter abstractly presents the steps that each computer program is going to take. In essence, chapter two is a study of the many components of each of the methods. The third chapter starts our analysis of these procedures. Two reference examples are given and a preliminary hypothesis is made. In the fourth chapter, some data on US Treasury STRIPS is looked at. both computationally and graphically. The final chapter presents a conclusion of all of these results.