|
|
BEATRIX JONES |
Inference for Parametric Models of Dispersal |
While a PMMB fellow in Elizabeth Thompsonís lab, I developed methodology for using spatial and genetic marker data in concert to infer dispersal models. The method is designed to look at dispersal on the time scale of one to a few years, using data from each reproductive season. In the modeling step, parametric models are used to keep dimensionality low and the models interpretable. Inference then is done in a maximum likelihood framework, using Markov Chain Monte Carlo methods to sum over the many possible assignments of offspring to parents admitted by the genetic data. |
My current research continues to utilize Markov Chain Monte Carlo methodology, but in a new context, using Gaussian graphical models to represent the covariance structure of gene expression levels, as measured in microarray experiments. Graphical models essentially represent network structures, so they are ideal for trying to infer regulatory relationships. In this context, MCMC algorithms are used to explore the space of possible networks. |
|
|
