Conceiving and Reasoning Visualization Design Options - Data Visualization: A Successful Design Process - page 84

• The second act refers to being able to satisfactorily judge the relative order or
ranking of values in terms of their magnitude. This is basic pattern matching
where we seek to determine the general hierarchy of the values being
displayed: where is the most and where is the least, which is the biggest and
which is the smallest.
• The lowest-level act relates to judging values. Studies have shown how the
effectiveness of different visual variables can be ranked based on which most
accurately support comparison and pattern perception. Bertin was the first
to propose such a hierarchy and his work has been tested, developed, and
refined by Cleveland and McGill ( Journal of the American Statistical Association,
Vol. 79, No. 387. September, 1984, pp. 531-554 ) and then by Jock MacKinlay.
Here, in the following presentation, we see the most recent version created by
MacKinlay. Each column represents the three main data types (note that there is no
distinction between ratio and interval-scale types of quantitative variables). Within
each column you have an ordering of the most accurate and least accurate visual
variables according to their interpretive precision:
Image recreated from "Ranking of Perceptual Tasks" ( Automating the Design of Graphical Presentations of
Relational Information, ACM Transactions on Graphics, Vol.5, No.2, April 1986 ) by Jock MacKinlay.
The studies by Bertin, Cleveland and McGill, and then MacKinlay focus on the
fact that our visual system isn't capable of absolute measurements. Therefore,
frameworks like this simply propose a guide to understand which variables will be
better at delivering relative measurements but with highest accuracy. In other words,
the higher up the column the easier it will be for your reader to accurately interpret
values represented by those variables.
So, looking at that table, you might ask why you would ever not use position as the
visual variable for your data: That will surely maximize the efficiency and accuracy
of your data communication for all data types?