[seqfan] A071816, A005900

[rhhardin at att.net]

Generalizing
http://www.research.att.com/~njas/sequences/A005900  (order 2)
http://www.research.att.com/~njas/sequences/A071816  (order 3)

numerically out to order 50, the ones I could compute enough terms for to check turned
out to be odd polynomials in n of degree 2*order-1.

The defining problem I used, eg., for order 8, just to choose one, was

a(n) = Number of ways the component sums of two 1..n 8-vectors can be equal

The polynomial for order 8 is empirically

a(n) = (2330931341/6810804000)*n^15 + (56057/340200)*n^13
+ (404711/3402000)*n^11 + (227197/2381400)*n^9 + (546533/6804000)*n^7
+ (130231/1871100)*n^5 + (266681/4299750)*n^3 + (1/15)*n

Transposing the resulting table a(order,n) there are two existing series

http://www.research.att.com/~njas/sequences/A082758 (n=3)
http://www.research.att.com/~njas/sequences/A005721 (n=4)

with the result that, eg., with the obvious generalization,

a(order,n) = largest coefficient of (1+...+x(n-1)) ^ (2*order)

and the curious question, is it obvious that this largest coefficient is an odd polynomial in n of degree 2*order-1.

I hope I've done variable name substitutions correctly above.









--
rhhardin at mindspring.com
rhhardin at att.net (either)