Bivariate Cauchy Regression
Document Type
Presentation Abstract
Presentation Date
9-14-2000
Abstract
For this distribution the moments or integrals do not exist that allow classical estimation of parameters. So one can revert to reliable nonparametric methods that are distribution free over the whole class of bivariate distributions that are elliptically symmetric; this includes the normal distribution and all the Student t distributions. The Greatest Deviation Correlation Coefficient will be explained and used. Most real data has a number of questionable data points and a method of regression that works on the Cauchy Regression can then be reliably used on real data without the data analyst worrying so much about the effect of “outliers”. The robustness of the method will be demonstrated by an example. PowerPoint will be used to demonstrate the geometric method of defining the Greatest Deviation Correlation Coefficient. Quantile plots are explained because they are a necessary tool for this method allowing scale factors to be estimated.
Recommended Citation
Gideon, Professor Rudy, "Bivariate Cauchy Regression" (2000). Colloquia of the Department of Mathematical Sciences. 70.
https://scholarworks.umt.edu/mathcolloquia/70
Additional Details
Thursday, 14 September 2000
4:10 p.m. in Math 109
Coffee/treats at 3:30 p.m. Math 104 (Lounge)