“Maybe” Should Be a Choice in “Yes-No” Questionnaires

Abstract

By permitting respondents to answer “don't know” or “possibly” to queries otherwise intended to be “yes” or “no” questions, a new dimension of the interview becomes available for analysis. By coding “yes” and “no” as “yes, I have an opinion that I will express to you,” and “don't know” or “possibly” as “no, I do not have an opinion that I will express to you,” understanding the patterning of such responses becomes an interesting research question. We present an example of this approach from 538 interviews, question-frames about the domain “illness,” obtained in the homes of participants in an intervention program designed to reduce coronary heart disease, in central, rural Mississippi. The questionnaire was presented by four interviewers. Each participant was interviewed four times at six-month intervals, providing adequate time between interviews for reflection on the task. We use the individual differences model of multidimensional scaling to obtain weights for each consultant on each dimension of the group aggregate space. Subsequent analysis of these weights was (1) by general linear model analysis of variance and (2) examination of the pattern of adjusted means of dimension weights by risk factors and design factors. Results were surprising. The two-dimensional aggregate space developed from opinions vs. lack of expressed opinion on individual questions was interpretable as one cluster of symptoms that implied heart disease and two other structures that were vector-like in appearance. Extremely high F-values showed a reflexive effect; the interviewer was associated with several factors including risk status of the participants, suggesting negotiation of whether or not a participant would agree to express an opinion. There may have been a reflective effect with changing patterns developing over the course of the repeated interviews. Neither dimension was associated with the health-care seeking behavior of consulting a doctor.