Resolution:
## Figure 2.
Results of the multiple logistic regression analyses relating to a woman’s likelihood
of choosing the pill, patch, or ring to the perception statements about these methods. The models fit the data well (c-statistic was 0.795, 0.802, and 0.820 for the models
for the pill, patch, and ring, respectively, which was considered acceptable to excellent
discrimination). The likelihood of selecting the pill, patch, or ring is higher if
the odds ratio (OR) is > 1 or lower if the OR is < 1 for women who ‘strongly agree/agree’
(or ‘strongly disagree/disagree’) compared with women who indicated they had ‘no opinion’
or ‘do not know’ about the perception statement in relation to the method. Two-sided
95% confidence intervals (CIs) are based on asymptotic normality of the OR estimates.
The likelihood of selecting a method is higher (by a statistically significant margin
at 5% level) if the lower limit of the CI is > 1. Conversely, the likelihood of selecting
a method is lower (by a statistically significant margin at 5%) if the upper limit
of the CI is < 1. If the CI contains 1, women who ‘strongly agree/agree’ (or ‘strongly
disagree/disagree’) compared with those who had ‘no opinion’ or ‘do not know’ do not
differ by a statistically significant margin with respect to the probability of choosing
the method.
Egarter |