This is a simple linear fitting problem, you solve it this way:
M = [x(:),y(:),z(:)]; % creating a matrix of all your independent parameters giving you a
% vector-matrix-versio of your equation:
% t = M*abc(:);
abc = M\t;
You can generalize this for arbitrary fitting-functions, just expand M with columns for x^2, sin(w*y) and the like. Matlab has all sort of fitting functions more general and agile. Have a look at polyfit, lscov and others
HTH
