Carrying on from @Matt's comment above... Is it really going to be more convenient to assemble all the elements into one long vector at the calling level? That's where/how to make the design decision on which is the better/best calling syntax.
function [AR,DEC] = astrometry(x,y,A,B,C,D,E,F)
is convenient to write the function succinctly, but burdens the caller with all eight arguments every time. The 8-vector interface code would look something like Matt's start or...
function [AR,DEC] = astrometry(arg)
A=arg(3);B=arg(4);C=arg(5);
D=arg(6);E=arg(7);F=arg(8);
There are many variations on a theme, of course, besides...
function [AR,DEC] = astrometry(x,y,C1,C2)
DEC = C2(1)*x+C2(2)*y+C2(3);
which would package the two sets of coefficients as vectors. Of course, they could also be
function [AR,DEC] = astrometry(x,y,C1,C2)
DEC = C2.A*x+C2.B*y+C2.C;
that puts the coefficients into two struct variables, each with consistent coefficient names by term, just different struct. That could then be an array of two as C(1).A, C(2).A, ... etc., or a single struct C.A where the field for each is a 2-vector and used as C.A(1) for first and C.A(2) for second.
The choices are all up to you, but your use then will have to be consistent to whatever choice you make.
