variation is largely consistent with results from principal components analysis (PCA) (Fig. 4 b). We note that the relationship between variancePartition and PCA depends on both the fraction of expression variation explained by a particular variable across all genes as well as the dimension of the variable. While variation across the 2 cell types explains less expression variation than variation across the 6 batches, the first principal component separates samples by cell type because this variable spans a lower-dimensional space.
