Abstract
This paper presents a number of algorithms for a recently developed measure of space-time concordance. Based on a spatially explicit version of Kendall's τ the original implementation of the concordance measure relied on a brute force O(n2) algorithm which has limited its use to modest sized problems. Several new algorithms have been devised which move this run time to O(n log(n) +np) where p is the expected number of spatial neighbors for each unit. Comparative timing of these alternative implementations reveals dramatic efficiency gains in moving away from the brute force algorithms. A tree-based implementation of the spatial concordance is also found to dominate a merge sort implementation.
Original language | English (US) |
---|---|
Pages (from-to) | 799-811 |
Number of pages | 13 |
Journal | Computational Statistics |
Volume | 29 |
Issue number | 3-4 |
DOIs | |
State | Published - Jun 2014 |
Keywords
- Autocorrelation
- Rank correlation
- Spatial concordance
ASJC Scopus subject areas
- Statistics and Probability
- Statistics, Probability and Uncertainty
- Computational Mathematics