* Run phase 1 once for each k
* Run for n = 50, 60, 70, ...

* Get shifted PMFs
* Get medians from shifted PMFs
* Check plots scaled by 1/median
* K-S test to see if these match one another
* Lin fit to get alpha and alpha*ln(k)
* From that, get k

* Then go back & check plots scaled by 1/(median*k)
* K-S test to see if these match the theoretical distribution ...

----------------------------------------------------------------

Theoretical distribution:
* c? alpha?

* google stable subordinator; stephen lalley?