|
Activity Number:
|
413
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Consulting
|
| Abstract - #310006 |
|
Title:
|
Bagging and Propensity Scores in Statistical Learning to Assign Ownership of Unclaimed Property
|
|
Author(s):
|
Marc Sobel*+ and Kenneth Swartz
|
|
Companies:
|
Temple University/Analysis and Inference, Inc. and Analysis and Inference, Inc.
|
|
Address:
|
Victoria Mills, Springfield, PA, 19064,
|
|
Keywords:
|
bagging ; propensity scores ; cart ; bootstrap ; payment reversal
|
|
Abstract:
|
Statistical methods have been used to estimate the total amount putatively owed by a private entity to a government entity under unclaimed property law. A private entity may receive payments without identification amounting to millions of dollars. Since investigations are time-consuming, and in many instances are impossible to complete, a statistical method was sought that would provide a sound basis for classifying any unidentified payment as a mistaken double payment (lawfully becoming unclaimed government agency property) or not (a legitimate payment for services rendered by the corporation). We employ the procedure of bagging classification and regression trees (CART) using bootstrap resampling to identify such double payments. Bagging techniques have achieved a great deal of success in many applications.
|
