Statistics seminar - Distinguished Professor Michael Stein - Rutgers New Brunswick

A seminar by Distinguished Professor Michael Stein from Rutgers New Brunswick

Title: Looking Carefully at Spatial Data

Abstract: As part of a book I am presently writing, I have been looking at elevation data over a 2000 x 2000 pixel region at 1/3 arcsecond resolution in the Colorado Rocky Mountains. I will discuss a number of issues with the data that I find statistically interesting and may also be of some scientific interest. One particularly notable aspect of the data is that the scale of local fluctuations can vary by orders of magnitude over distances as short as 100 m. This result calls in to question the standard approach in spatial statisics of handling nonstationarity by assuming the process is at least locally stationary. A second interesting finding concerns the evidence for spatial variation in the apparent smoothness of the process over the wholly mountainous region. The values of these local estimates of smoothness differ substantially between different estimators in ways that are difficult to explain. These results and others provide the basis for general discussion on the value of traditional geostatistical models for large spatial datasets.

For further information, please contact RSFAS Seminars.

All information collected by the University is governed by the ANU Privacy Policy.