|
Activity Number:
|
604
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #304791 |
|
Title:
|
Bootstrap Integration of Multiple Measurement Sources with Applications to Inference on Extremes
|
|
Author(s):
|
Russell L. Zaretzki*+
|
|
Companies:
|
The University of Tennessee
|
|
Address:
|
328 Stokely Management Center, Knoxville, TN, 97996,
|
|
Keywords:
|
booststrap ; risk analysis ; reliability ; extreme values
|
|
Abstract:
|
The development of cheap sensor technology has led to an increasingly frequent situation in risk analysis and reliability where a large database of sensor measurements is available along with sparse information on some outcome. For example, in auditing the cleanup of a radioactive waste site, exhaustive scan measurements of radiation levels are routinely collected across an entire parcel of land. A small secondary sample of accurate laboratory measurements is also collected. The analyst must integrate the scan data with the more accurate laboratory measurements for inference regarding the population parameters such as extreme percentiles of the distribution of radiation emission. We present a bootstrap method that integrates these measurements and show how it may be used for inference. We also discuss sample size requirements and improvements in accuracy resulting from the method.
|
