Consider two models predicted using tenfold cross-validation. The accuracy produced by model1 is 32, 15, 14, 20, 35, 45, 48, 52, 27, and 29, and model2 is 20, 14, 10, 8, 15, 20, 25, 17, 19, and 7. We want to determine which model performance is significantly better than the other at 0.05 significance level.

a) When do we use tenfold cross-validation technique?

b) Which statistical test will be used to compare the performance of two models? Justify your answer.