Figure 1.

Percent missingness type by volatile features (number of features of each missingness type divided by the total number of features), and percent contribution of each volatile feature’s missingness to the overall missingness profile (number of missing values within each feature for each missingness type divided by the total number of missing values). (A) Classification using Pearson correlation and the KS test. (B) Classification using Pearson correlation and Cucconi test. “Excluded” refers to features that could not be evaluated by the KS test because of high degrees of missingness (≥60%). MCAR was not detected in either data set.