|
Activity Number:
|
602
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #305136 |
|
Title:
|
The Modified Parametric Spatial Bootstrap
|
|
Author(s):
|
Jose H. Guardiola*+ and Hassan Elsalloukh
|
|
Companies:
|
Texas A&M University and University of Arkansas at Little Rock
|
|
Address:
|
6300 Ocean Drive, Corpus Christy, TX, 78412,
|
|
Keywords:
|
spectral decomposition ; spatial correlation ; resampling ; square root matrix ; confidence interval
|
|
Abstract:
|
The classic bootstrap method resamples observations that are assumed to be identically distributed and independent. When the classic bootstrap method is applied to spatially correlated data it destroys the correlation structure of the data. This paper proposes a modified method for performing a parametric bootstrap that produces valid resamples of spatially correlated data. Once that the corresponding confidence intervals are obtained, the percentage of confidence intervals that capture the true mean are computed. The coverage of confidence intervals produced by this method is compared to the parametric spatial bootstrap method proposed by Tang (2005). This study concludes that the proposed new procedure produces slightly better confidence intervals than the method proposed by Tang, but still needs a better computational algorithm that takes advantage of a sparse matrix structure.
|
