ABSEIR is an R package designed to efficiently implement a class of spatial epidemic models known as spatial SEIR(S) models. This family of analytical techniques is more completely described in this manuscript, as well as in my thesis work, though a brief practical introduction is also given here. Installation instructions for ABSEIR are available on the wiki, and any problems with the library or this series of tutorials can be directed to the issues page.
If you’re reading this, you’re probably interested in epidemic modeling techniques. It will therefore come as no surprise to you that epidemics develop over both time and space. When we want to apply statistical models to these phenomena, however, we need to be a bit more specific about what we mean when we say ‘time’ and ‘space’. The class of spatial SEIR(S) models is defined for discrete time points and discrete spatial locations. This may take the form of weekly reports on a particular disease process, perhaps subdivided by administrative regions. This specification stands in contrast to continuous spatiotemporal processes, like the latitude, longitude, and time at which a biological sample is taken.
The acronyms in this field can be a bit daunting, but allow readers to very quickly get an idea for what kind of assumptions a particular model is likely to make about a disease process. Disease states are broken up into ‘compartments’, and the names/labels of these compartments generally give rise to the acronyms. SEIR models, for example, describe models in which an individual may either be Susceptible, Exposed, Infectious, or Removed. Susceptible indiduals do not have a particular disease, but may contract it. Exposed individuals (somewhat confusingly) have contracted the pathogen of interest, but are not yet capable of spreading it. Infectious individuals are… infectious, and the removed category describes individuals who are no longer infectious due to recovery with immunity or mortality.
Clearly, these are not the only disease states we might be interested in. In fact, ABSEIR already allows users to fit full SEIRS models, which allow removed individuals to re-enter the susceptible population. More generally, we often want to distinguish between various types of infectious state, different removal states (with different probabilities of reentry into the susceptible population), work with multiple species or make any number of other generalizations to the compartmental framework. Work in this area is ongoing, though the ABSEIR package itself is likely to retain a broadly similar feature set to that it currently possesses. There are certainly a lot of exciting opportunities for the development of more general epidemic modeling software. Even so, you might be very surprised at the wide range of analyses which can be performed using the spatial SEIR(S) model class as implemented by ABSEIR. We’re capable of working with a variety of spatial and non-spatial models for a wide variety of disease processes. As we’ll see in later tutorials, the ‘spatial’ structure as described in the above manuscripts can be adapted to describe structured multi-host populations.
The ABSEIR documentation is intended to give an informal introduction to important concepts for understanding the compartmental models implemented by the software, and to describe the practical details of actually working with these models. It is not intended to provide a comprehensive background on epidemic modeling techniques or approximate Bayesian methodology more generally. Questions are, however, encouraged, and should be directed to the issues section of our GitHub page. This documentation also does not cover installation, Detailed instructions are available here.