A large current archeological work study the space occupation evolution in time by the societies. The observation of these spatial transformations can depend on a geostatistical approach through global and local Exploratory Spatial Data Analysis (ESDA). The map is a unique exploratory tool. In fact, exploratory spatial data analysis methods provide a way of observing the spatial organizations. By combining these calculation methods, it is clearly possible to bring out spatial structures or put forward new hypotheses. These approaches allow identifying any possible spatial particularities or local clusters at different step of time (Cressie 1993, Fotheringham et al 2000). They are used to create in fine map of change.

However, questions remain about how to evaluate the reliability or robustness of these hypotheses and these results? How to compare these spatial forms over time? How to interpret these maps of change? To extend that, firstly, the “statistical sample” used is based on archaeological information including complex inventory properties (heterogeneity, incompleteness, data missing) are not measurable. On the other hand, statistical processes tested in the methods used (K Ripley) are based on homogeneous spatial phenomena, which is not appropriateness to archaeological data. Indeed, Orton has underlined that “their use [of K-Ripley statistics] is based on the assumption of a homogeneous and isotropic point process; there may well be reasons in practice why such an assumption does not hold” (Orton 2004, p. 303), suggesting the Pélissier and Goreaud’s approach (Goreaud, Pélissier 2001) wich consists to dissect spatial patterns into different scale.

As, archaeologists deal with vast inventories, the recurring question is how to use them? How to incorporate this uncertainty in data analysis?

An experience could be carried out with very different geographically and chronologically sets of archeological data from several programs. This experience inspired in part the work of Bevan and Conolly (2006). They are the pioneers to show that the robustness of statistical results could be validated with application of multi-scalar and sampling approaches for each study area. These methods seem to overcome some of the problems of missing data. Finally, one of the perspectives of this experiment is the use in archeology from the comparison of seed points to non-homogeneous point process as developed and described these recent years in the field of ecology (Perry, Miller 2006). These methods are used to explore the soundness of spatial pattern detection for archaeological inventories. A modified form of the K-function suitable for inhomogeneous point pattern is used (Baddeley, Turner 2005), which takes into account the variation of pattern’s intensity in space.