Fig. 5 Change in correct ordering under five-subset minimum-size matching. The heatmaps show all 64 unequal-size, nonzero-information-difference cells for permutation-centred KSG with k = 4 in the moderate-imbalance extension. Panels a-d identify the population mutual information I_A and I_B in nats. The smaller sample has N_A = 100 or 200 observations; N_B equals the plotted ratio times N_A. L1 is the (1,1) Gaussian design; Q1x is the (4,3) quadratic design with one active coordinate pair. Matching retains A and averages estimates from five size-N_A subsets of B, sampled without replacement within each subset; subsets can overlap. Each cell gives the paired matching-minus-baseline change in the percentage of 601 independent parent pairs with correct ordering. Positive values indicate improved ordering and negative values indicate worse ordering. Parent catalogues are reused across anchors, ratios and truth comparisons, so the 64 cells are dependent. The values are point estimates, without significance symbols; their paired pointwise 95% bootstrap intervals do not all exclude zero and have no simultaneous-coverage guarantee. The accompanying Figure5_plotted_values.csv and the complete extension tables in Online Resource 2 provide uncertainty intervals, ordering probabilities, comparison RMSE and absolute RMSE
