Activity Number:
|
181
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 1, 2016 : 10:30 AM to 12:20 PM
|
Sponsor:
|
IMS
|
Abstract #318906
|
|
Title:
|
A Matching Coalescent with Application to Testing Model Adequacy with Heteroscedastic Variances
|
Author(s):
|
James Neill* and Forrest Miller
|
Companies:
|
Kansas State University and Kansas State University
|
Keywords:
|
Lack of Fit Tests ;
Matching Coalescent ;
Heteroscedastic Variances
|
Abstract:
|
We give a hierarchical type algorithm which determines a unique clustering of n data points in p-dimensional space based on a sequence of weighted graphs. At each stage we obtain a minimal weighted maximal matching which determines the next graph and a new partition. The algorithm merges more than just two clusters at each stage using a maximal matching. We use a maximum method to define the weights for the new graphs. The process stops when there are no usable edges left and gives the final partition. The algorithm is computationally feasible for large data sets and the number of clusters need not be specified beforehand. We discuss the use of these methods in testing regression function adequacy in the presence of heteroscedastic variances.
|
Authors who are presenting talks have a * after their name.