4.6 Data robustness

In a regional investigation, the numerical data on the artefact composition of sites should be reduced to a lower and more stable level of measurement. First of all a reduction to an ordinal level can be considered. Instead of observing that in site 52B-168, 14% of macrolithic artefacts, 29% of scrapers and 57% other artefacts occur, a more reliable observation is that there is more production debris than scrapers and more scrapers than macrolithic artefacts. In contrast with the percentages, these data will in general be affected less by, for example, new field surveys. Yet doubts can be raised whether those ordinal data - macrolithic artefacts(1), scrapers(2), other artefacts(3) - may be considered stable enough.

By a reduction to the nominal level of measurement the data might be made more 'robust': macrolithic artefacts, scrapers and production debris are present, whereas borers, pottery and grinding stones are absent. This approach does not seem desirable on the core regional level. Nominal data are least subject to change, to be sure, but this makes a site-typological analysis less illuminating. We did choose this option for the macro-regional level, especially in view of the much more limited quality of the primary data.