A Q-mode cluster analysis of the percent-transformed data, which clusters samples based on the percentages of contained species, set reveals that samples from similar paleoenvironments (as determined by sediment type) tend to cluster closer to each other than to samples taken from other paleoenvironments, a rather obvious result.
An R-mode analysis, which clusters the species based on the samples in which they are contained, was somewhat more interesting. Very few species clustered together at a very high level of similarity (i.e. occurring together at a high level of frequency). This indicates that even though the samples are similar enough to each other to cluster into reasonable groups, the species do not occur only with a well-defined group of other species, but commonly occur with different species in different samples. Therefore, while the analyses revealed structure in the composition of the samples, the species are not strictly limited to appearing in only certain paleoenvironments, but can occur in low abundances in any paleoenvironment. This is probably due to the presence to the patchiness of the paleoenvironment in both space and time.
A simple principle components analysis of the samples in the data set required 7 principle axes in order to explain 95% of the variability contained in the matrix, once again indicating a high degree of variability in the data matrix. When the results of the analysis were plotted in the PCA space defined by the first three axes, three loose groupings were revealed. When the samples from each section were connected in stratigraphic sequence, all of the sections from the eastern part of the collection area traced the same basic path through factor space, despite the high variability of the data matrix. (Fig. 1) The collections from Petersborough (LTR collection locality), near the western edge of the outcrop area, followed a different path through the defined space.
Figure 1: Principle components analysis of all Yorktown samples. The principle components space is defined by three axes. Each colored line records the path of each individual section through factor space. The lowest stratigraphic sample is connected to next in stratigraphic order in sequence until all samples have been connected. Muddiness (a paleoenvironmental indicator) is also mapped onto the PCA analysis. A factor loading analysis reveals that these axes are defined based on:
The similar paths traced by the eastern samples indicates that the localities within that portion of the depositional basin experienced the same basic changes in faunal composition and relative abundance during the changing paleoenvironmental conditions of that time. Even though the species contained within each paleocommunity are highly variable, the gradients from one "end-member" paleocommunity to another is discernible with the data available.
Back