Anandamayee Majumdar, PhD
Assistant Professor, San Francisco State University
This research studies spatial data for multivariate processes. Specifically, it develops a covariance structure for multivariate spatial variables that are heterogeneous extensions of homogeneous processes. Using kernel-convolution techniques, it generalizes the kernel-convolution method developed by Majumdar and Gelfand (2003) to processes where there are local centers of stationarity. Simulations show that the exact knowledge of local centers of stationarity is not necessary to develop such models. This research has been applied to soil pollution data where local centers of stationarity can be initially chosen from the land-use types (urban, rural, agricultural, and more). Spatial statistics and analysis have been widely applied in biostatistics, genomics, and epidemiological studies to study disease outbreaks.