PREVIOUS   NEXT   CONTENTS   ISSUE   HOME 

8.1 Multivariate analysis

We have seen that geographical space works well as a frame of reference when we are looking into the problem of temporal development of the Holešov cemetery. But taken alone it is virtually useless (at least at this site) once we want to gain some knowledge of the details of funerary rites, social structure and similar issues. We can only use the same dataset we already have, so what needs to be changed is the frame of reference. An ideal solution is to perform a multivariate synthesis of the data and search for patterns.

A technical note on the terminology used here is probably needed before we look at the analysis itself: Neustupný (1997) prefers the term multivariate (vector) synthesis and not analysis, which is more frequent in the literature. An 'analysis' means a deconstruction of a context into its basic elements and in archaeology this is usually realised by visual description and/or laboratory analyses. A synthesis is yet another step in the process of scientific reasoning, where isolated facts produced by the analysis are combined into patterns. This leads to the recognition of meaningful structures that are the subject of interpretation (Neustupný 1993). It may then be difficult to keep terminology exact when speaking about methods where the somewhat inappropriate term analysis is deeply rooted (e.g. trend surface analysis, cluster analysis, principal component analysis and so on).

To summarise my approach in the Holešov case-study, I carried out correlation analysis on eleven (most frequent) attributes of the descriptive database (Table 2), principal components analysis (PCA), and factor analysis with the Varimax rotation of extracted factors (Neustupný 1997). Six factors, or main axes of data variability, were extracted from the dataset (Table 3 and Table 4). The first four have properties pointing at their strong affiliation to specific demographic groups and cemetery zones, which can reflect both gender affiliations and chronological structures. The other axes of variability show much weaker relations in respect of gender and temporal dimensions and obviously require other interpretations (Šmejda 2003b). Focus will now turn to the first four factors, which are most pertinent to the research goals set for this study.