(Images generated by SDM)

By changing the size or shape of sets of objects so that they are distinct from others, retinal shift operations can also be used to classify objects and eliminate occlusion while maintaining context. Both of these goals may be met by the figures below. In contrast to the original figure (above), in the left figure the widths of selected objects are expanded by a constant amount so that they are larger and more noticeable, and unselected objects are shrunk and made thin. In the figure on the right, the selected bars are differentiated first by their color and then further distinguished by elongating their shape. Users can define and change their own groupings or classifications throughout the data analysis process.

It is important to note that shift operations do not maintain relative size or height ratios among objects. Thus, changing object parameters that are tied to particular data attribute values means that we can no longer compare the relative heights or sizes of the altered objects, although simple differences can still be observed. In the dataset shown in this figure, the widths of objects are not used to show any data attribute, so no information is lost.

Go to next page.