An example of randomly missing data imputation. Here, the left-most pane shows the inferred mean function for the gene g
(t), and the next three panes show three replications, in the last of which the missing data imputation occurs. Other replications are omitted for brevity. The final panel shows a zoom of the imputed region (boxed in the previous panel). Imputation by the HGP is marked by a circle on the mean function; replicate median is denoted by a square; PCA is denoted by a diamond and KNN by a triangle. The HGP imputation is closer to the replicable statistic, being less affected by cross-sectional noise effects than the other methods.