A machine learning practitioner uses 5x2 cross-validation to compare two models. Which statistical test is specifically designed for this comparison procedure?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — you want Dietterich's 5x2 CV paired t-test because it's built for the 5x2 design and handles the variance and correlation structure. An independent t-test on all fold results ignores the paired nature of the folds.
Full Explanation
Dietterich's 5x2 CV paired t-test is correct because it accounts for the variance and correlation structure of the 5x2 design. The independent samples t-test on all 10 fold results ignores the paired nature of CV folds. McNemar's test applied to the confusion matrices of both models tests differences on a single test set. The Wilcoxon signed-rank test on 5 fold differences from a single 5-fold run describes a different approach, not the 5x2 design.