G*Power [47] was used to calculate sample sizes to detect a mode bias of 1.5 in population prevalence using McNemar’s Exact Test for paired nominal data, assuming 25% of participants would switch indicator status (above or below threshold) between rounds.