Graphics = The Human Eye + Data + Theory

                    John Maindonald, ANU

The human eye/brain combination is the instrument that reads meaning into 
graphs. Data come both in the form of data for plotting, and experimental 
data on the processes of human perception. Theory comes both in the form 
of statistical theory that gives insights on what to plot, and a theory 
of human perception that indicates that some forms of presentation will 
be effective, and others ineffective. Graphs may fail both because they 
reveal too little, and because they lie and thereby reveal too much. In 
practical data analysis, graphical presentation goes hand in hand with 
model-based analysis. Both for graphical presentation and for analysis, 
parsimony and hierarchy help cope with complexity. Parsimony can be 
overdone, and checks are essential. This talk will explore these ideas 
in the context of multiple regression, discriminant analysis (here 
seen primarily as a dimension reduction technique) and principal 
components analysis.