1.  The dataset PLAY4.SYD contains two measurements (X, Y) on 17 objects (A-Q).  Your assignment is to do a cluster analysis of these objects by hand.  The clustering procedure should be polythetic agglomeration using euclidean distance.  Join clusters by means of single linkage or complete linkage, as you prefer (but be sure to be explicit about the method you choose).  More specifically, here's what you should do:

• First, make a scatter diagram on a piece of paper and measure euclidean distance with a ruler.  From this produce a dendrogram.  Also do a line graph showing dissimilarity versus the number of clusters.
• Are there "natural" clusters within this dataset?  If so, at what level of clustering do these occur?  Justify your answer with reference to the dendrogam, scatter diagram, and line graph(s) you have drawn.

2.  Now use the data in PLAY4.SYD to do some clustering with SYSTAT's hierarchical clustering procedure (click on Statistics | Classification | Hierarchical Clustering).  Using euclidean distance, cluster the objects (rows) by means of single linkage, complete linkage, average linkage, and Ward's method.  Compare and discuss your results.

3.  Now let's get back to the real world.  The file ARCHAIC.SYD contains data on assemblage composition from 10 Early Archaic sites on the Atlantic slope of the eastern U.S. (the data are taken, slightly modified with Randy Daniel's help, from the references listed on the reverse).  The variables in the file are the following: SITE$, site name; HBIFACE, hafted biface count; OBIFACE, other biface count; FUNIFACE, formal uniface count; EUNIFACE, expedient uniface count; CTOOL, cobble tool count.  Your job is to analyze the variability in assemblage composition and to tell me something (brief) about what it might mean.  (You should feel free to refer to the articles from which the data are taken, but don't feel compelled to read them extensively.  The main thing is to look at Anderson and Hansen's cluster analysis on pp. 278-280.)

• Convert the counts to percentages and cluster the sites using a variety of methods.  How do your result's compare with Anderson and Hansen's?  How do results of the different clustering methods compare to one another?
• Now standardize the percentages and cluster the data again, using the same range of methods as before.  Discuss your results.

References for part 3:

Anderson, David G., and Joseph Schuldenrein (1983).  Early Archaic settlement on the southeastern Atlantic slope: A view from the Rucker's Bottom site, Elbert County, Georgia.  North American Archaeologist 4(3): 177-210.

Anderson, David G., and Glen T. Hanson (1988).  Early Archaic settlement in the southeastern United States: A case study from the Savannah River valley.  American Antiquity 53(2): 262-286.

Datasets for this exercise (right-click to download):

• Play4.syd (delete ".txt" suffix after downloading)
• Archaic.syd (delete ".txt" suffix after downloading)

Back