Skip to main content
. Author manuscript; available in PMC: 2011 Jun 1.
Published in final edited form as: Pattern Recognit. 2010 Jun 1;43(6):2340–2350. doi: 10.1016/j.patcog.2009.12.003

Fig. 6.

Fig. 6

(a) Effect of the number of subsets on the performance of the method for the human brain image database system (HBIDS). This figure compares the performance for 10%, 20%, and 30% missing values. As a rule of thumb, 6 to 10 subsets are sufficient. Let alone how much data is missing, more than 10 subsets do not seem to improve the performance. (b) Receiver Operating Characteristic (ROC) curves for the surgery candidate selection problem. The selection-fusion (SF) and Multiple Imputation (MI) methods are compared. Note the superiority of SF.