Fig. 4
From: Machine learning identifies a compact gene set for monitoring the circadian clock in human blood
Predicting CT using various numbers of control samples from the same individual. Boxplots of absolute error from tenfold cross-validation. From left to right, the numbers of groups (one prediction per group) were 498, 198, 200, and 121. Levels of statistical significance (Wilcoxon rank-sum test): * P = 2.5 × 10–4, ** P = 1.9 × 10–5, *** P = 5.9 × 10–7. Other differences were not statistically significant