Data Spatial Analysis in Archaeology
Anthropology 589b Fraser D. Neiman 2.15.07 University of Virginia Spring 2007
Problem Set 3: House Size Variation at AR5
The Problem Liz Arkush has provided is with an extraordinary set of data on the sizes of houses and their locations on within a fortified, hilltop settlement site AR5 in Peru. What can spatial variation in house size tell us about the social lives of their inhabitants? Some specific questions to ponder include: Was there social inequality in this community? How was inequality distributed in space? Was the settlement comprised of a single social hierarchy, or is there evidence for multiple groups? 1 The tools you have to answer this question include trend-surface analysis, loess, and correlograms. And don't forget looking at the raw data a site plan from ArcGIS with house size variation shown. To make your lives easier, I have written SAS code that fits the trend and loess surfaces, computes residuals, maps the trends and residuals, and computes correlograms for the residuals well as as the raw diameter values. There are two code files: AR5TrendSurface.sas and AR5Loess.sas on the class website. Experiment with them and compare results. How do the trend surfaces compare with the loess fits? Which explains more variation remember our friend "r-squared" from the last class? What do the residuals do? Which lag size allows you so see pattering in autocorrelation the best? In thinking about the results, a critical question is: to what extent does decomposing the spatial variation in house diameter into a large-scale trend and local, spatially-dependent variation in the residuals ...