Create a small dataset of no more than 20 (maybe even 10) samples with 3 clusters. Manually calculate each of the cvi values for this dataset for each cvi and assert that the output of each cvi function matches the manual calculation exactly. This ensures that no matter what future changes we make, the cvis are still working exactly as intended and makes for simpler debugging.