Report 4

Author

012345678

Data Sources

The rating data are from the Psychological Science Accelerator project: To Which World Regions Does the Valence-Dominance Model of Social Perception Apply? (Jones et al. 2021).

The face stimuli are from Ma, Correll, and Wittenbrink (2015). While the original image set refers to “target self-identified race (A = asian; B = black; L = latinx; W = white)”, we have already renamed this variable to “Ethnicity” and refer to it as such throughout the report.

The 120 faces were rated on a 1-9 scale for 13 different characteristics: “aggressive”, “attractive”, “caring”, “confident”, “dominant”, “emostable”, “intelligent”, “mean”, “responsible”, “sociable”, “trustworthy”, “unhappy”, and “weird”. Data are presented for average ratings in each of 11 world regions.

Trust and Dominance

Factor analysis revealed two factors, labelled “valence” and “dominance”. Trustworthiness is often used as a single rating proxy for valence, and dominant for dominance.

Figure 1 shows the distribution of trustworthiness and dominance for each region.

Figure 1: The distribution of ratings for trustworthiness and dominance in the European regions

UK Ratings

Table 1 shows the mean and standard deviation of trustworthiness and dominance ratings for the UK, broken down by gender of face.

Table 1: Mean and SD of trust and dominance ratings for the UK by gender.

Trustworthy

Dominant

Gender

mean

SD

mean

SD

Female

5.50

0.73

4.46

0.55

Male

4.90

0.67

4.46

0.61

Analysis

In this section, I will analyse the UK ratings using two-tailed t-tests with a critical alpha of 0.05 to test for gender differences in trustworthiness and dominance ratings.

In the UK, there was a significant gender difference in trustworthiness ratings (t(117) = 4.67, p < .001, 95% CI = [0.34, 0.85]) and a non-significant gender difference in dominance ratings (t(117) = 0.05, p = 0.958, 95% CI = [-0.20, 0.22]).

Simulation

I simulated a new dataset for the UK from Table 1, creating a new dataset of average trustworthiness ratings for a simulated sample of 30 faces of each gender (see Figure 2).

Figure 2: Simulated trustworthiness and dominance ratings for the UK. Error bars show the 95% CI around the mean.

Iteration and Power

Since only trustworthiness showed a significant gender difference, I then used simulation to calculate power for the effect size found in the UK data for a sample of 5 to 60 faces (in steps of 5) of each gender, just for trustworthiness.

Figure 3: Show the relationship between N and power in a plot with an appropriate caption.

Reflections

Reproducibility

  1. Thing I did

    How it improves reproducibility…

  2. Thing I did

    How it improves reproducibility…

  3. Thing I did

    How it improves reproducibility…

Sources of Learning

Please explain the sources of learning you used in this assessment, including the book, help sessions, peers, online sources, and generative AI. (up to 300 words)

References

Jones, Benedict C, Lisa M DeBruine, Jessica K Flake, Marco Tullio Liuzza, Jan Antfolk, Nwadiogo C Arinze, Izuchukwu LG Ndukaihe, et al. 2021. “To Which World Regions Does the Valence–Dominance Model of Social Perception Apply?” Nature Human Behaviour 5 (1): 159–69.
Ma, Debbie S, Joshua Correll, and Bernd Wittenbrink. 2015. “The Chicago Face Database: A Free Stimulus Set of Faces and Norming Data.” Behavior Research Methods 47: 1122–35.