Graphical Lyapunov models can be used to analyze dependencies among the variables of random vectors. In the continuous case we assume that the observations arise from the equilibrium distribution of an Ornstein-Uhlenbeck process. This distribution is determined by two unknown parameter matrices: by the drift matrix, which is assumed to be sparse, and by the volatility matrix, which is assumed to be diagonal. The drift matrix captures information about cause-effect relations among the variables of the observations. In the first part of this thesis we explore two methods based on lasso regression to recover the support of the drift matrix using the data and we compare their performance through simulations. In the second part we ask under which assumptions it is possible to identify the parameter matrices if the support of the drift matrix and the covariance matrix of the distribution are known. We provide general conditions for identifiability and we give examples of identifiable supports.     «

Graphical Lyapunov models can be used to analyze dependencies among the variables of random vectors. In the continuous case we assume that the observations arise from the equilibrium distribution of an Ornstein-Uhlenbeck process. This distribution is determined by two unknown parameter matrices: by the drift matrix, which is assumed to be sparse, and by the volatility matrix, which is assumed to be diagonal. The drift matrix captures information about cause-effect relations among the variables o...     »