A "doublely-recursive" sequence

[Edwin Clark]

> 
> Hmmm...I also wonder what the plot of this sequence, perhaps scaled 
> logarithmically, looks like.
> 

I plotted the points (n,floor(log[10](a[n]))) for n from 1 to 2000.

I then plotted the points (n, rand(1..floor(sqrt(n))) ) for n from 1 to
2000. 

Here, of course, rand(1..floor(sqrt(n))) is a pseudo random integer in the
range 1 to floor(sqrt(n)).


The two plots look quite similar.  

---Edwin Clark