Events per variable for risk differences and relative risks using pseudo-observations