Figure 3From: Cross-validated stepwise regression for identification of novel non-nucleoside reverse transcriptase inhibitor resistance associated mutationsReference and 3F methodology: schematic overview. In the 3F method cross-validated prediction error (CVPRESS) was used instead of significance levels (p-values) in the reference approach. In the reference approach two stepwise regression procedures were used: all possible mutation pairs were made from mutations in the first order model and candidate for entry in the second order model. In the 3F method, the initial search space consisted of all individual mutations. Mutation pairs could only enter the model if both mutations in the pair were already selected for the model.Back to article page