Loading Events
  • This event has passed.

Re-Framing I: Correcting the Scaling of Spatial Autocorrelation and a New Method to Tackle Big Data

March 22, 2018
3:00 pm - 4:00 pm

Event Navigation

Dr. Thomas J. DeWitt
Department of Wildlife & Fisheries
Associate Professor

Moran’s I is an iconic measure of spatial autocorrelation. It is elegant for its intended basic and intuitive nature, ofen idealized beyond its actual character. Several authors have noticed problems of bias, distribution assumption violations, and scale exception snafus with the original statistic. Still, I is generally treated as its Platonic ideal. Ideally, I ranges with a bell­‐curve form between -‐1 and 1, where its value indicates perfect dispersion at ­‐1, random interspersion at 0, and perfect clustering at 1. The intuitive and pedagogic value of this ideal are important to protect if I can be modified to actually achieve the interpretations just described. Another practical limit in the use of I is its development for low­‐dimensional applications. A single environmental variable or linear combination of data generally is tested for autocorrelation. Modern data sets typically involve two or more large, internally-­consistent data blocks. Thus, N-­block approaches must be developed. I present my solutions for these issues and end with a universal N-‐block version of I that is applicable to all data structures. This new approach is illustrated with population and ecogeographical examples.

About Dr. DeWitt

Dr. DeWitt obtained graduate and postdoctoral training at Boston University, Binghamton University (SUNY), University of Michigan, and University of Kentucky. He studies ecological mechanisms that drive evolutionary diversification in space and time. He has also published on statistics, biomathematics, paleontology, and paleoarchaelogy. These studies involve a diversity of aquatic organisms such as fish and snails, although through collaborations has published on plants, microbes, insects, and mammals such as seals, otters, and primates. He has published in Science and PNAS and high­‐impact disciplinary journals such as American Naturalist, Evolution, and Ecosphere. His metrics include several thousand citations and an Erdös number of 5.

Contact: Michael Bishop at michael.bishop@tamu.edu

Details

Date:
March 22, 2018
Time:
3:00 pm - 4:00 pm

Venue

Eller O&M Bldg., Room 807