Statistical similarity