ABSTRACT
Graphics are effective ways of summarising and conveying information. Once the various graphics forms have been described, the textbooks can pass on to supposedly more difficult topics such as proving the central limit theorem or the asymptotic normality of maximum likelihood estimates. The evidence of how graphics are used in practice suggests that they need more attention than a cursory introduction backed up by a few examples. Graphics make these points directly and give an overview that is easier to remember than sets of numbers. Graphics are good for qualitative conclusions and often that is what is primarily wanted. Graphics are for revealing structure rather than details, for highlighting big differences rather than identifying subtle distinctions. Graphical Data Analysis is obviously appropriate for observational data where the standard statistical assumptions that are needed for model building may not hold. For some of the graphics the code includes adjustments to improve the look of the default versions.
