Clustering of Pilots

We performed a complete-linkage clustering on the matrix of inter-subject correlations. We considered solutions with from 2 to five clusters. The four cluster solution yielded clusters of reasonable sizes (305, 105, 84 and 53 subjects each). The clusters are plotted in the space defined by the principal components analysis in figures 10 thru 12.

Cluster 1, the largest cluster is low on PC1 (discomfort). These pilots are comfortable with the automation. Cluster 2 is higher on PC1 and very high on PC2(Magic). These pilots also tend to have less time than the pilots in cluster 1.

None of the observed clusters maps cleanly onto PC3 (by the book).