The ordered nature of the regression justifies the use of the map as a display for data items. When the items are mapped to those units on the map that have the closest reference vectors, nearby units will have similar data items mapped onto them. Such an ordered display of the data items facilitates understanding of the structures in the data set. Kohonen (1981) was the first to propose using such displays to illustrate a data set.
The same display can be used for displaying several other kinds of information. One clear advantage of always using the same display is that as the analysts grow more familiar with the map, they can interpret new information displayed on it faster and more easily.
For example, the map display can be used as an ordered groundwork on which the original data variables, components of the data vectors, can be displayed in their natural order. Such displays have been demonstrated in Publication 2. The variables become smoothed locally on the display, which helps in gaining insight in the distributions of their values in the data set. Such displays are much more illustrative than, for instance, raw linearly organized statistical tables. It might also be useful to display the residuals, average differences of the variables from their smoothed values.