Geostatistics in heterogeneous regions with cost-based distances. In a sentence, I am trying to find valid covariance functions that depend isotropically on a non-Euclidean distance. To the right, you can see (and download) the poster presented at the IWSM2011 conference, with a brief statement of the problem and an overview of my lines of work. Here
is the final dissertation. It is in spanish only, sorry. Although you
can read the abstract in english. I hope to translate the key
chapters soon. Defense took place feb. 15th, 2013. Here are the slides of the lecture. |
This was the motivating applied problem for my Thesis
project: You can get access to the dataset, code and a demo script here. This project motivated the following article. |
I have also been doing some research on spatio-temporal models for disease mapping. Related to this, I knew about INLA, and got very interested in its application to Bayesian inference in latent Gaussian models.
I am also very interested in the ongoing development of the SPDE
approach for geostatistical inference, which is connected with INLA, and very
related with my thesis project.
In late 2011, I have been attending the SAMSI research Program on Uncertainty Quantification.
In a google
code repository there is a script that installs and configures
several programs and tools that me and my colleagues use for
statistical research. It is to be used in a Ubuntu (or Mint) GNU/Linux
machine. The software includes LaTeX, R, several editors and plugins,
the Bazaar VCS, and some GIS repositories. It is intended to set up a
machine and leave it up and running.