2. 장소 : 통계학과 스마트강의실 (자연대연구실험동 222호)

 

3. 연사 : 김대영 교수 (University of Massachusetts Amherst, Department of Mathematics and Statistics)

 

4. 연제 : Association Learning for Multidimensional Categorical Data with Ordinal Outcomes

 

Abstract

 

 

The analysis of multivariate categorical data with ordinal responses is an important task in various science fields.
An essential preliminary step of effectively analyzing this data type is the exploration of potential dependency structures among the variables in a flexible and unstructured manner.
This exploratory phase can uncover dependency patterns  that might have been only informally acknowledged or unrecognized,  as well as identify variables that could impact the ordinal response variable.
In this talk we introduce a recently developed data-driven and model-free methodology designed to explore regression associations  in  multivariate data with an ordinal outcome and categorical  (nominal/ordinal) predictors.
The proposed method involves the development of  the checkerboard copula regression and its corresponding association measure,  which aids in identifying, visualizing and quantifying  regression associations between the ordinal outcome  and a set of categorical predictors of interest in an integrated manner.
We will discuss the theoretical properties of the proposed methods and demonstrate their performance with a real data set.