A comparison of bootstrap methods and an adjusted bootstrap ...
A comparison of bootstrap methods and an adjusted bootstrap ...
A comparison of bootstrap methods and an adjusted bootstrap ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
BIASSTDMSECase 10.10.050-0.05-0.1-0.15-0.2-0.25-0.3-0.35OBSBCV0.632LOOCVOOBLOOBS0.632+ABS0.20.150.10.050OBSBCV0.632LOOCVOOBLOOBS0.632+ABS0.120.10.080.060.040.020OBSBCV0.632LOOCVOOBLOOBS0.632+ABSBIASSTDMSECase 20.020-0.02-0.04-0.06-0.08-0.1-0.12-0.14-0.16OBSBCV0.632LOOCVOOBLOOBS0.632+ABS0.120.10.080.060.040.020OBSBCV0.632LOOCVOOBLOOBS0.632+ABS0.0250.020.0150.010.0050OBSBCV0.632LOOCVOOBLOOBS0.632+ABSFigure 3. Comparison <strong>of</strong> Prediction Error Estimation on Simulated Datasets with n=40, p=800. Case 1: Nodifferentially expressed genes; Case 2: Class 0 patients follow normal distribution with me<strong>an</strong> 0, for class 1patients, the 2% differentially expressed genes follow normal with me<strong>an</strong> 1.5. Classification <strong><strong>an</strong>d</strong> regression tree isused in class prediction. The “true” prediction errors for the two cases are 0.500 <strong><strong>an</strong>d</strong> 0.221. Methods displayed areordinary <strong>bootstrap</strong> (OBS), <strong>bootstrap</strong> cross-validation (BCV), 632 <strong>bootstrap</strong>, leave-one-out cross-validation (LOOCV), out<strong>of</strong>-bagestimation (OOB), leave-one-out <strong>bootstrap</strong> (LOOBS),.632+ <strong>bootstrap</strong>, <strong><strong>an</strong>d</strong> <strong>adjusted</strong> <strong>bootstrap</strong> (ABS).