askazy, you driving me crazy with the Wilcoxon. Try to add which test in particular, since his name is associated with many tests.

Yes, you could subtract the two paired values and use the Wilcoxon sign rank test (on the differences). What is your sample size and data distribution. Many programs will conduct an exact test when the sample sizes are small (e.g., in SAS, n-value < 20), also some people feel the central limit theorem comes into play as the sample size increases.

This is a better dataset idea than the previous proposed sets!