Figure 4.

Estimates of topology accuracy (%) for neighbor-joining clusters of four samples representing four groups in each SNP genotype data set (A: corn, B: rice, and C: wheat) over 100 runs of simulation with 10–90% of total observations as random missing (M, missing data without imputation) and as imputed with an imputation method (RF, random forest; PP, probabilistic principal component analysis [PCA]; or NI, nonlinear iterative partial least squares PCA).