Objectives: To determine the effects of 5 different numerical response scales--continuous-open, dichotomous-yes/no, and continuous-closed numerical (CCN) with 3 different ranges of response frequencies (low, medium, high)--on the proportion of respondents defined as regular exercisers.
Methods: We randomly assigned 500 undergraduate students to complete 1 of the 5 numerical response scales.
Results: The percentage of participants defined as regular exercisers ranged from 14% in the CCN low-frequency group to 45% in the CCN high-frequency group [chi2 (4, 500) = 28.90; P<.001].
Conclusions: The different numerical response scales have a significant impact on the estimated percentage of regular exercisers.