On possibilities and limitations of using self-anchoring scales in web surveys