Stable and Efficient Computation for Population-Level Integral Projection Models

• Activity Number: 569
• Type: Invited
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Alan E. Gelfand*+
• Companies: Duke University
• Keywords: Cox process ; hierarchical model ; plant demography ; shot noise process

Abstract

Atmospheric science and, more generally, environmental science is often concerned with how changes in climate will affect the behavior/performance of species. Here, we focus on tree species and, in particular, on response to environment in the form of trait distributions (such as size, age, mass) for trees in a stand. We work with demographic projection using integral projection models. Fitting these models is very challenging, impossible to do without approximation, except for models too simple to be of interest. A further challenge is to scale these models from stands to large regions (e.g., the eastern U.S.). Another challenge is to work with data having extreme levels of missingness, more than 80% in our database. So, after providing sufficient introduction to the problem, we present various computational strategies which enable feasible, stable, and efficient computation in the face of these challenges.