This book is about graphical methods for multivariate data, and their uses in understanding relationships particularly when there are several aspects to be considered together in multiple response models such as multivariate analysis of variance and multivariate multiple regression.
It brings together a collection of novel techniques I, and others, have developed over the past 15 years and implemented in mature R packages. Some key substantive features of the book are:
-
Statistical data visualization is cast in a general framework by goal (see the data, visualize a model, diagnose problems), rather than a categorization by graphic types. It is best informed by principles and goals of communication, for example making graphic comparison easy and ordering factors and variables according to what should be seen (effect ordering).
-
Data visualization is seen as a combination of exposure---plotting the raw data---and summarization--- plotting statistical summaries---to highlight what should be noticed. For example, data ellipses and confidence ellipses are widely used as simple, effective summaries of data and fitted model parameters. When the data is complex, the idea of visual thinning can be used to balance the tradeoff.
-
These graphical tools can be used to understand or explain a wide variety of statistical concepts, phenomena, and paradoxes such as Simpson's paradox, effects of measurement error, and so forth.
-
The HE ("hypothesis - error") plot framework provides a simple way to understand the results of statistical tests and the relations among response outcomes in the multivariate linear model.
-
Dimension reduction techniques such as PCA and discriminant analysis are presented as "multivariate juicers," able to squeeze the important information in high-dimensional data into informative two-dimensional views.