Figure 1: A. Distribution of misophonia severity scores across all participants (n = 676). B. Distribution of different trigger types across all participants (n = 676).
Figure 2A shows the relation between the likelihood of mimicry and misophonia severity using univariate logistic regression. The probability to mimic increased as a function of misophonia severity (X2 (1, n = 676) = 47.42, p < .001; Figure 2A). The model explained 9% (Nagelkerke R2) of the variance and correctly classified 59.8% of cases, with an odds ratio (OR) of 1.219.
After confirming the relation between mimicry and overall misophonia severity we explored whether mimicking was associated with triggers of a particular type. Multivariable logistic regression was performed, with mimicking as a response variable and sensitivity on seven different sound categories as explanatory variables. Results of this model are shown in Table 1. Overall, the model could successfully predict the outcome in 59% of cases (X2 (1, n = 676) = 40.86, p < .001) and explained 7.8% (Nagelkerke R2) of the variance. Sensitivity for eating sounds was the greatest predictor of mimicry (OR = 1.341; p = .001) followed by sensitivity for consonant/vowel sounds (OR = 1.154; p = 0.046). No other sound category had a significant relationship with mimicking. Plots of odds ratio for each sound category with 95% confidence intervals are shown in Figure 2B.