It would seem that you can, but what is your goal?
Would truncating the table serve just as well, or will the recreated table have a different structure?
What do you want to happen when the table does not yet exist? (Is that even possible?) Handle creation in this same changeset, or in a separate one?
As for using a single change set, what if your DBMS doesn’t support DDL transactions (e.g., MySQL), and the drop succeeds, but then the recreation somehow fails? Will you be able to recover from the intermediate state where the table is missing? (That’s one reason in favor of using a separate change set for each atomic change.)
What people will usually want is to first use one changeset to drop the table (or rename it, in case you later need to roll back), subject to a precondition that it already exists, then another changeset to create the new table.