Fig. 4From: Combining location-and-scale batch effect adjustment with data cleaning by latent factor adjustmentVisualization of the assimilation of validation batch to training batch after batch effect adjustment. First two principal components out of PCA performed on the following data matrix: the training batch after batch effect adjustment combined with the validation batch after addon batch effect adjustment. The training batch in each subplot is depicted in bold and the numbers distinguish the two classes “IUGR yes” (2) vs. “IUGR no” (1). The contour lines represent batch-wise two-dimensional kernel estimates and the diamonds represent the batch-wise centers of gravities of the pointsBack to article page