Activity Number:
|
69
- Statistical Methods in Ecology
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, August 7, 2022 : 4:00 PM to 5:50 PM
|
Sponsor:
|
Section on Statistics and the Environment
|
Abstract #322067
|
|
Title:
|
Generalized Linear Network Models with Application to Multi-Species Count Distributions from the ForestGEO Database
|
Author(s):
|
Nels Johnson* and Susan Cordell and Christian Giardina and Amanda Uowolo and Inyoung Kim
|
Companies:
|
USDA Forest Service, Pacific Southwest Research Station and USDA Forest Service, Institute of Pacific Island Forestry and USDA Forest Service, Institute of Pacific Island Forestry and USDA Forest Service, Institute of Pacific Island Forestry and Virginia Tech
|
Keywords:
|
Markov field;
pairwise process;
point process models;
species distribution model;
species interactions
|
Abstract:
|
We show how to build several multivariate exponential family distributions with dependent dimensions. These exponential family distributions are equivalent to Markov random fields or pairwise processes. To illustrate these techniques, we implement Poisson and negative-binomial versions of the model on ForestGEO tree count data from the Republic of Palau. The implementation is fully Bayesian and uses Hamiltonian Monte Carlo to sample from the posterior. We also perform a simulation study to assess the robustness of model selection techniques on this kind of data.
|
Authors who are presenting talks have a * after their name.