Bill almost has the right idea. Unfortunately, tracing descendants on the woman's side from the 1700's to find descendants to test would be extremely difficult, and then true descendants 8 generations from an ancestor might not have got any DNA from the ancestor, and certainly would be unlikely to share anything with a specific 8th cousin.
But the idea of the wives is correct. You might be able to make use of a clustering tool. There are many available that put your matches together in groups that likely correspond to common ancestors.
Do this clustering for each of the Levi descendants. Then compare the matches in common between the descendants and see what cluster they are in. In a perfect world you'll have a cluster for Samuel, Joseph and Abraham, another for Joseph's wife and a third for Abraham's wife.
Of course, it's not a perfect world, and there could be endogamy involved, which will complicate matters. But even if the clusters are not clear, this exercise should give you some insight that might be helpful.