CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1210
Title: Efficient model search over DAG causal hypotheses about brain connectivity within a given population Authors:  James Quartermaine Smith - University of Warwick (United Kingdom) [presenting]
Abstract: Directed DAGs of probabilistic dynamic processes are currently now being used as frameworks for describing the dynamic relationships between centres of activity in a given unit. For example one recent application of these methods can be found in applications of Markov graphics to model he dynamically evolving relationships between the regions of a brain - as measured by fMRI experiments - as a particular individual engages in various activities. The space of models of this type is obviously huge and so scoring the efficacy of different explanations in terms of their fit to this highly multivariate time series data is a challenging one. However an important related issue is how inferences can be made concerning the nature of the connectivities between centres of activity lying in subpopulations of individuals - for example those suffering from a particular medical condition as opposed to those who do not. The aim is to discuss some new efficient search methodologies - originally designed for the analysis of models concerning single subjects - can be adapted for use in the study of these even larger composite systems. We will then discuss some of the inferential and interpretational challenges these methodologies still face.