Sure they're vectors.
(scroll to see the rest of the output)
Of course, IA is right. If v is very long at all, the problem becomes impractical. If the range of values in v allows, you might be able to save some weight by using uint8(). Still, even when using uint8(v), the result for a vector of length 25 occupies 4.2GB in RAM.
For what it's worth, we can save some time (might be valuable if n gets larger than about 20) by actually calculating how big C needs to be. I just let the array grow last time, but that was just me being lazy.
s = factorial(n)./(factorial((1:n)).*factorial(n-(1:n)));
C(sc(k)+(1:s(k))) = num2cell(c,2);
The time savings are maybe 10-20%, though that probably varies with version, hardware, and environment.