A scale that allows a respondent to express relative magnitude between the answers to a question and arranges responses in a hierarchical pattern, but does not allow researchers to determine absolute difference between responses, is called a(n) .