Fig. 5: Rank-derived scores of synthesis models and metric correlations.
From: A Multifaceted benchmarking of synthetic electronic health record generation models
a Rank-derived scores on synthetic data generated using UW data. b Rank-derived scores on synthetic data generated using VUMC data. c A heatmap of the Pearson correlation coefficients for pairwise metrics on the rank-derived scores across all candidate models for the two real datasets (UW and VUMC). The best (i.e., lowest rank) and worst scores for each metric are in bold red and black font, respectively.
