Abstract
This thesis aims at presenting a set of concepts and inferential tools adapted to the analysis of neutral genetic polymorphism in a specific class of stochastic population genetics models : spatialized demographic models of populations where the dispersal of individuals between generations is limited in space (isolation by distance models). Such analysis is based on the combination of stochastic models of population evolution (generation by generation coalescence) and statistical inference method by simulation (“the summary-likelihood method”). In a first step various tools were developed and tested independently before being integrated together to create an inferential framework adapted to our models. In a second step, we used this framework to study the performances and benefits of these new inference methods and the influence of considering linkage disequilibrium patterns on parameter inference under distance isolation models.