Fig. 3

Benchmarking for the Mixture of Gaussians Data Set. The top panel includes the observations as simulated, colored according to the mixture component. The bottom panel presents the adjusted Rand index as a function of the missing data proportion for 8 different approaches to handling missing data; a higher value indicates better agreement between the predicted and true cluster assignments, adjusting for chance. Error bars represent the standard error of the mean across 20 simulation replicates