Page 61 - Demo

59How attractiveness preferences influence attention3as a reference point to calculate the looking time bias, by calculating thebias toward the left picture. Hereafter, we have tested whether this bias isaffected by (1) the attractiveness ratings of the left and right picture, and(2) date outcome.To study the association between attractiveness and voluntary attention,we modelled Looking time bias score (proportion of time looking at the leftpicture) as a function of Attractiveness rating of the left picture and Attractiveness rating of the right picture, and their interactions with Gender.We allowed the intercept to vary by Subject. Importantly, we weighed eachtrial by the looking time in that trial relative to the subject’s average (seeData Processing). Thus, trials in which the participant paid more attentionto the screen had a larger weight in the analysis. In this manner, we avoidedthat trials where participants were distracted or disinterested would have alarge influence on the outcome of our analysis. Furthermore, we specifiedthe same formulas for the precision parameter (phi; shape of the beta distribution), the zero-one inflation parameter (zoi; probability of observing azero or a one), and the conditional one-inflation parameter (coi; probabilityof observing a one if a zero or one is observed). To study the associationbetween date outcome (i.e., willingness to go on another date with datingpartner) and voluntary attention, we followed the same procedure as described above. However, the predictors Attractiveness rating of the rightpicture and Attractiveness rating of the left picture were replaced with Dateagain right picture (binary: yes/no) and Date again left picture (binary:yes/no).We used a Gaussian prior with M = 0 and SD = 0.25 for the Interceptof the beta component of the model. For the independent variables, wespecified regularizing Gaussian priors with M = 0 and SD = 0.5. This alsoapplied to the independent variables in the formulas for phi, coi, and zoi.For all variance parameters, we kept the default Student’s t priors with 3degrees of freedom. Furthermore, we kept the default logistic priors for theIntercepts of zoi and coi, and default Student’s t prior with 3 degrees offreedom for the Intercept of phi.After running the models, we used the emmeans-package (Lenth et al.,2023) to integrate the different model components, and provide estimatesbased on the posterior predictive distribution. Using these values, we calculated multiple quantitative measures to describe the effects (see StatisticalAnalyses). It is important to note, though, that the predictions are on theresponse scale (probability). This complicates interpretation for the continuous variables, because the slope on the response scale is not constant butis shallower or steeper depending on the value of the continuous variable. Inthe text we report the effect size measures for when the continuous variableof interest is set at 0, but in the Supplementary Material we provide similarmeasures for other values of the continuous variable of interest.Iliana Samara 17x24.indd 59 08-04-2024 16:35