Figure 2

Canonical variable plots of control liver data. Each panel shows the first two canonical variables, which represent the maximum achievable separation for each factor (in rows) and array type (column), computed from the first ten principal components. Each point in the plot represents an individual sample. The marker color along with text indicates the factor level for the sample and the shape indicates the site where the data were generated. Coded study is indicated by site number followed by a letter indicating study within a site.