In the last decade, graphical models have become increasingly popular as a statistical tool. This book is the first which provides an account of graphical models for multivariate complex normal distributions. Beginning with an introduction to the multivariate complex normal distribution, the authors develop the marginal and conditional distributions of random vectors and matrices. Then they introduce complex MANOVA models and parameter estimation and hypothesis testing for these models. After introducing undirected graphs, they then develop the theory of complex normal graphical models including the maximum likelihood estimation of the concentration matrix and hypothesis testing of conditional independence.

The present set of notes grew out of our interest in the study of statistical transformation models, in particular exponential transfor- mation models. The latter class comprises as special cases all fully tractable models for mUltivariate normal observations. The theory of decomposition and invariance of measures provides essential tools for the study of transformation models. While the major aspects of that theory are treated in a number of mathematical monographs, mostly as part of much broader contexts, we have found no single account in the literature which is sufficiently comprehensive for statistical pur- poses. This volume aims to fill the gap and to indicate the usefulness of measure decomposition and invariance theory for the methodology of statistical transformation models. In the course of the work with these notes we have benefitted much from discussions with steen Arne Andersson, J0rgen Hoffmann-J0rgensen and J0rgen Granfeldt Petersen. We are also very indebted to Jette Ham- borg and Oddbj0rg Wethelund for their eminent secretarial assistance.