How can it be?

The Bayesian estimates honor, in a statistically defensible fashion, 1) sample size in a given quadrat, 2) mean similarity of a quadrat’s type frequencies to the average value for the neighborhood, and 3) mean uncertainty about type frequencies within quadrats in a neighborhood. Bayesian estimates are, therefore, superior to current methods that rely on simple weighted moving averages (e.g., Neiman 1990, Whallon 1984). Spatial variation in artifact-type frequencies likely is caused by both temporal and social variation. Common practice in archaeological spatial analysis, based on cluster analysis, confounds these dimensions of variation. CA offers a means to disentangle them.