Modeling non-Gaussian processes on a sphere using multi-resolution analysis
Debashis Paul (University of California)
Friday 16th March, 2018 14:00-15:00 Maths 116
We propose a model for random processes on a sphere having non-Gaussian and scale-dependent characteristics. The construction of the process is based on a multi-resolution frame called needlets. We implement an MCMC scheme for inferring the model parameters and for spatial prediction. We illustrate the methodology through the analysis of a data set obtained through simulations of high-latitude ionospheric potential fields. We also discussion potential extensions to non-Gaussian vectorial processes on the sphere.
This is a joint work with Minjie Fan, Thomas Lee and Tomoko Matsuo.