Note that the diagonal lines in this plot are actually not symmetric—they intersect at a risk allele frequency less than 0.5, and reflections neither vertically nor horizontally will make them match. We may have assumed that there should be symmetry, for example by interchanging the cases and controls to switch between dominant and recessive model. However, this is not valid since they are ascertained differently, the controls being a sample from the whole population and the cases from the diseased subset.
