Table 7 Diabetes dataset covariates identified as correlated with principal components. Results of a search for covariates with value distributions having significant differences in either UP _n vs. DOWN _n , {UP _n +FLAT _n } vs. DOWN _n , or UP _n vs. {FLAT _n +DOWN _n } are shown below. An "X" indicates the covariate's values varied significantly in the corresponding principal component's sample partitioning. 11 (out of 53) covariates were identified at a 1% significance level (minSetSize = 5) in 10 of the first 20 principal components. To investigate further, corresponding plots of covariate value distributions within sample partitions can be generated, such as the covariate plots for PC₁₄ (see Figure 5).

From: Mining gene expression data by interpreting principal components