EEMS uses a population genetic model that involves migration on an undirected graph G = (V,E) with vertices (demes) V connected by edges E. The graph G is a regular triangular grid, which is fixed and embedded in a two-dimensional plane, so that each deme has a known location and only neighboring demes are directly connected (Figure 1b ). The density of the grid is pre-specified by the user and depends on both computational considerations – computational complexity scales cubically with the number of vertices – and the resolution of the available spatial data.
The EEMS model has migration parameters m and diversity parameters q, where m = {me: e ∈E} specifies an effective migration rate on every edge and q = {qv: v ∈V} specifies an effective diversity rate for every deme. Intuitively, the migration rates m characterize the genetic dissimilarities between distinct demes, while the diversity rates q characterize the genetic dissimilarities between distinct individuals from the same deme. The EEMS model is a special case of the general stepping stone model 28 (link), which allows directed migration as well as migration between demes that are not located close in space.
We use Bayesian inference to estimate the EEMS parameters m and q. Its key components are the likelihood, which measures how well the parameters explain the observed data, and the prior, which captures the expectation that m and q have some spatial structure (in particular, the idea that nearby edges will tend to have similar migration rates).
The EEMS model has migration parameters m and diversity parameters q, where m = {me: e ∈E} specifies an effective migration rate on every edge and q = {qv: v ∈V} specifies an effective diversity rate for every deme. Intuitively, the migration rates m characterize the genetic dissimilarities between distinct demes, while the diversity rates q characterize the genetic dissimilarities between distinct individuals from the same deme. The EEMS model is a special case of the general stepping stone model 28 (link), which allows directed migration as well as migration between demes that are not located close in space.
We use Bayesian inference to estimate the EEMS parameters m and q. Its key components are the likelihood, which measures how well the parameters explain the observed data, and the prior, which captures the expectation that m and q have some spatial structure (in particular, the idea that nearby edges will tend to have similar migration rates).
