The more flexible continuation ratio model was preferred to a binomial regression because of the variability between questions in the proportion of 'research oriented' answers.
Here, we consider the multivariate version of this problem: having n units which are classified according to q binary questions, we want to summarize the association between questions in terms of quasi-implications between features.
A statistical significant difference in the CTT difficulty indices (U = 21, P-value <0.05) and IRT difficulty indices (U = 13, P-value <0.05) was found between questions in the knowledge and comprehension levels.
Differences between questions in the databases were present (Table 1).
One patient expressed a concern with repeating herself and there were no indications of similarity between questions in the other DT interviews.
We interpreted this as evidence that the questions were equally difficult, and therefore made no distinction between questions in subsequent analysis.
For axial coding, concepts were organized into overarching themes, and compared, both within and between questions, in search of patterns in responses and to ensure consistency.
However, the more striking result is the difference between scores on the analogous questions in a pair (e.g., #5 vs #10 and #17 vs 6; Figure 4b and 4d, respectively) where a respondent in one question (e.g., #5) is asked to indicate what the "scientific community" believes, and in a second question (e.g., #10) what the student believes.
The men surveyed were aged between 17 and 74 and answered questions in a computer-based self-interview.
Most nominees answer "between 60 and 70 percent of their questions in a fully forthcoming manner," the study, by Dion Farganis of Elon University in North Carolina and Justin Wedeking of the University of Kentucky, concluded.
