GRANDPA: GeneRAtive network sampling using degree and property augmentation applied to the analysis of partially confidential healthcare networks

Applied Network Science

Table 2 Comparison of network measures between the original CMS subgraph and various generated graphs

	Degree		Eigenvector centrality		Community agreement
	NRMSE	KS (p value)	NRMSE	KS p value	Community agreement
PGM (specialty)	0.404	0.387 (< 0.001)	0.167	0.901 (< 0.001)	0.039
PGM (specialty and degree)	0.045	0.069 (0.109)	0.167	0.873 (< 0.001)	0.039
GRANDPA (specialty and community)	0.237	0.248 (< 0.001)	0.165	0.543 (< 0.001)	1.000
GRANDPA (specialty, community and degree)	0.054	0.038 (0.777)	0.088	0.173 (< 0.001)	0.993
GRANDPA (specialty, community, degree and linchpin)	0.051	0.033 (0.897)	0.028	0.102 (0.003)	0.990

Normalized root mean square error (NRMSE) and Kolmogorov–Smirnov (KS) test p values were calculated to evaluate the distribution of degree and centrality measures across the generated graphs
Propensity matching at the vertex level was conducted between the original and generated graphs to calculate community agreement