SEMINARS |
|
Fall 2006
STATISTICS
COLLOQUIUM
Wednesday, November 29, 2006
3:30-4:00—Refreshments, Yost 327
4:00-5:00—Talk,
Yost, Room 101
Tonglin Zhang, PhD
Department of Statistics
Purdue University
Spatial Clustering Detection in Loglinear Models
Spatial clustered data are common and many spatial phenomena, such as births, deaths, crimes and species richness, can be counted by a spatial unit, either as a raw count or as a ratio over some exposure. Not like the Gaussian data, the heterogeneity problem is a serious concern in the detection of spatial clustering in the count data. In this talk, I am going to display a solution to the heterogeneity problem in the loglinear model framework and display a revision of a few well known test statistics for spatial clustering detection. The simulation results show that the behavior of the type I error probabilities and the power functions are much better than those of the corresponding test statistics without adjusting for heterogeneity. In addition, by carefully defining a local association term, the proposed method is able to also identify a local cluster as well as its shape and size.
|
| |
