I am going to carry out a small scale action research on ratio problems. I would ask the same questions in different ways and see whether wordings would affect students' performance. However, past research showed that numerical values would affect questions' difficulty. Say comparing (1:2 and 1:4) and (3:7 and 5:11), the latter one is obviously more difficult.
How can I control for the numerical factors so that I can make inferences on "problem wordings"?