Abstract
A graphical model is a probability model for multivariate random observations whose independence structure can be characterised by independence graph, and graphical modelling is the statistical activity of fitting graphical models to the data. The theory of graphical modelling has emerged from a mixture of log-linear models in multi-way contingency tables and covariance selection models in multivariate normal distribution. In this article, we develop the conversational data analysis system for graphical Gaussian modelling which can be used for analysis of multi-dimensional continuous variables when assuming their distribution to be the multivariate normal distribution. This system is applicable to representation of a probability model by using not only undirected independence graphs but also directed and chain independence graphs. The system can be linked to the existing analysis system for other methods of multivariate analysis. The effectiveness is illustrated through an example.
