Scaling regimes of 2d turbulence with power-law stirring: theories versus numerical experiments