Page 66 - Demo

64Chapter 3Figure 3. Effect plot showing associations between Pre-date attractiveness ratingand Looking time bias separate per Gender. The black line represents the medianeffect, while the grey ribbon represents the 95% credible interval.right picture (bwomen−men = 0.012 [.010], 89% CrI [-0.006; 0.032], pd+= 0.91).This pattern was similar for other values of Pre-date attractiveness rating:there was no robust difference in slope between men and women (Table S15).We performed the same analysis with the Post-date attractiveness ratings as predictor. This analysis yielded the same results (Table S16). Altogether, the results show that participants indeed looked longer at faces thatthey rated as attractive. The results are visualized in Figure 3.Date outcomeSecond, we investigated the association between Date outcome and Look-ingtime bias using Bayesian zero-one inflated beta regression (Descriptives:Table S17-19; Model Table: Table S20). We found that participants showedmore attention towards pictures of people that they later indicated theywould like to date again. More specifically, when the left picture depictedsomeone they wanted to date again, they spent on average 12.4 percentagepoints longer looking at the left picture than when the left picture depictedsomeone they did not want to date again (bno−yes = -0.124 [.019], 89% CrI[-0.154, -0.095], pd−= 1.00). When the right picture depicted someone theywanted to date again, they spent on average 15.8 percentage points less looking at the left picture than when the right picture depicted someone they didnot want to date again (bno−yes = 0.158 [.017], 89% CrI [0.131, 0.186],pd+ = 1.00).Figure 4. Plot showing the effect of Date outcome on Looking time biasseparate per Gender. Error bars represent 95% Credible Intervals.To see whether the effect was modulated by Gender, we investigatedwhether the effect for women and men was substantially d ifferent. However,we found no consistent gender differences (Left p icture: b women−men = 0.060[.037], 89% CrI [0.002, 0.118], pd+= 0.95; Right picture: b women−men = 0.014[.034], 89% CrI [-0.043, 0.066], pd+= 0.66), although the pd suggested thatthe effect of Date outcome on Looking time bias was stronger for men forthe left picture specifically.Altogether, the results show that participants indeed looked longerat the faces of people that they later indicated they wanted to see againafter their speed-date (Figure 4).Model comparisonsRegarding the immediate attention analysis, we found no clear differences inpredictive accuracy between the three models (Table S21-23). Although themodel that included Pre-date attractiveness rating had the highest ex-pectedlog-predictive density, the differences with the models that included Postdate attractiveness rating (∆elpdLOO = 10.5 [9.6]) or Date outcome(∆e lpdLOO = 14.0 [11.2]) as predictors was not robust due to the relativelyhigh standard errors. Thus, while the model that incorporated Pre-date attractiveness rating as predictors had the highest predictive accuracy, therewas no substantial difference in predictive accuracy with the two other models.Figure 4. Plot showing the effect of Date outcome on Looking time biasseparate per Gender. Error bars represent 95% Credible Intervals.To see whether the effect was modulated by Gender, we investigatedwhether the effect for women and men was substantially different. However,we found no consistent gender differences (Left picture: b women−men = 0.060[.037], 89% CrI [0.002, 0.118], pd+= 0.95; Right picture: b women−men = 0.014[.034], 89% CrI [-0.043, 0.066], pd+= 0.66), although the pd suggested thatthe effect of Date outcome on Looking time bias was stronger for men for theleft picture specifically.Altogether, the results show that participants indeed looked longerat the faces of people that they later indicated they wanted to see againafter their speed-date (Figure 4).Model comparisonsRegarding the immediate attention analysis, we found no clear differences inpredictive accuracy between the three models (Table S21-23). Although themodel that included Pre-date attractiveness rating had the highest ex-pectedlog-predictive density, the differences with the models that included Postdate attractiveness rating (∆elpdLOO = 10.5 [9.6]) or Date outcome(∆e lpdLOO = 14.0 [11.2]) as predictors was not robust due to the relativelyhigh standard errors. Thus, while the model that incorporated Pre-date attractiveness rating as predictors had the highest predictive accuracy, therewas no substantial difference in predictive accuracy with the two other models.Iliana Samara 17x24.indd 64 08-04-2024 16:35