Graphical models are important statistical tools used for exploring the relationship between the variables of a multivariate distribution. They are regularly used for analysing medical data and are a well studied problem in biostatistics. However, the majority of the models are intended for the case where all variables are either discrete or continuous. Very few methods exist for finding the relationship between ordinal and continuous variables which is a common scenario in practice. For instance, in medical studies, researchers take a series of measurements (continuous variables) from patients along with a questionnaire with multi-answer questions (ordinal variables) answered by the patients themselves to study the cause of a certain disease. Exploring the statistical relations between the measurements and the questions is an essential element for their research. Hence, there is an important demand for models which can be used on data consisting of both continuous and categorical variables. There has been recent advances in the study of high-dimensional graphical models for mixed data offering practical solutions to this problem. In this thesis, we take the latest contributions to this field and engage in an empirical comparison study under various scenarios to test the limits of the most recent methods. Our goal is to derive recommendations for practitioners who wish to use these methods on real data.     «

Graphical models are important statistical tools used for exploring the relationship between the variables of a multivariate distribution. They are regularly used for analysing medical data and are a well studied problem in biostatistics. However, the majority of the models are intended for the case where all variables are either discrete or continuous. Very few methods exist for finding the relationship between ordinal and continuous variables which is a common scenario in practice. For instanc...     »