On the representation of major stratospheric warmings in reanalyses