5.3 Segment diagrams

Let's display the various segment diagrams of the 23 'clean' Michelsberg sites. The consequences of opting for absolute numbers, percentages or ordinal classes per artefact type become clear now, as the different diagrams are displayed one above the other.

Of the Michelsberg sites the following characteristics have been displayed in a segment diagram and visually interpreted:

• overall size. The circle increases with an increasing number of finds per site. Methodologically it might be argued that such a circle diagram could be considered a special kind of segment diagram, viz. where only a single variable is plotted. The picture shows there is one very sizeable site, one sizeable site and a large group of small sites.
• range of artefact types. The circle increases with an increasing range of artefact types. The picture shows the two large sites have a wide range of artefact types and the small sites lag far behind in range of artefacts.
• amount of artefacts of various types. In 15 segment parts the 'raw' numbers of each artefact category have been plotted. The first segment represents the number of triangular arrowheads, the last segment the number of other artefacts. In most sites this last segment is predominant. The numbers of the other artefact types are not or barely visible, 'outshone' as it were by the large numbers of production debris.
• percentage of the various artefact types. The percentage segments provide a much better image of the occurrence of the various artefact types. This is particularly true when the absolute numbers are low, but that is precisely when percentages are not very reliable.
• ranking the various artefact types. Here the artefact types have been listed in order of absolute numbers. Two or more artefact types may have the same frequency, a so-called tie. These artefact types have equal rank then, and have the value of the lowest in the group (minimum tie). For example there are six artefact types on site 52E-150 with no finds. They are all awarded a ranking of 1. The first find category with a single find is the group of triangular arrowheads with invasive surface retouch, which is awarded a ranking of 7. The highest ranking number is always 15. Quite often for ties the average of the rankings is used (average tie). The six categories without any finds would then all receive a ranking of (1+2+3+4+5+6)/6 = 3.5. Visually however this is less clear, so a minimum tie solution is preferred. The picture of the sites clearly shows the differences: there are two sites with a large range of artefacts, beside a large number of sites consisting of only one or two types.
• progressive class values of the various artefact types. For this, the progressive class values (1:accidental, 2:very few, 3:few, ranging to 10:abundant) have been plotted for each artefact type in the segments. In ranking, the categories without any finds still received a value of 1, but in this type of diagram these find categories are invisible. Artefact types with an almost identical frequency now have the same class value, whereas they had consecutive ranking in the previous situation. This series provides a good picture of both the range of artefacts (2 varied sites), and the absolute size (2 large, 3 intermediate and many small sites).

The various graphic series display the information in different ways; we prefer the last series of the class segments, particularly because here a great deal of information from the other diagrams appears to be combined. We have therefore chosen to use segment diagrams in the Meuse Valley Project, where progressive ordinal classes are plotted. These offer us, we feel, a good way of presenting multivariate data, keeping in mind the limited quality of the data as well.