Convection in a box: on the dependence of preferred wave-number upon the Rayleigh number at finite amplitude