Chapter 4: Results
4.3. Objective 2
4.3.1. Question 3
Does the sex of an individual determine how attractive or masculine they find certain male faces and body odours?
Visual Stimulus Attractiveness
The Friedmanโs two-way ANOVA comparing rankings of stimuli items between the sexes was not significant both before a bootstrap was applied (ฯยฒ= 7, df =5, NS) as well as after (๐๐ฅ2=5.37, df=5, NS). Therefore, the null hypothesis was not rejected and the conclusion that there was no significant difference between the ranked stimuli items, drawn. However, as there were no significant differences between the ranked items it can be assumed that men and women ranked the items differently.
Although the Friedmanโs test implies a difference between the average rank of the medians between men and women, a ๐ฬ test indicates significant agreement both before and after a bootstrap for both the sexes combined. However, although the agreement is significant it is small, and furthermore the ๐ฬ statistic decreases after the bootstrap. In addition, the maximum agreement estimated by the upper confidence limit suggests that we can be 95%
certain that the ๐ฬ population will not be greater than 0.286, which indicates a small yet significant window of agreement. Men in the initial sample did not significantly agree on which of the VS they considered visually attractive, however after a bootstrap was applied the men showed small but significant agreement. Women showed moderate significant concordance regarding the VS. This result may support the Friedmanโs test. However, after a bootstrap was applied with the sample size of the sexes equated both men and women, independent of each other, showed agreement in their rankings of VA. However, in this instance women still showed more agreement regarding rankings of VA. Please see table 15 for a summary of the ๐ฬ findings.
Table 15: Summary of bootstrapped (B=10000) ๐พฬ for rankings of visual attractiveness for the sexes
Sex ๐ฬ P ๐๐ฬ P SE Lower
confidence limit
Upper confidence
limit Decision Both
sexes
combined 0.18 <0.001 0.143 <0.001 0.062 0.045 0.284 Reject Women 0.33 <0.001 0.362 <0.001 0.116 0.142 0.597 Reject
Men 0.063 NS 0.107 0.05 0.062 0.026 0.266 Reject
The histograms below indicate the distributions of ๐ฝ๐พฬ for all significant aggreements in rankings of VA. The first histogram indicates the distribution of ๐ฝ๐พฬ statistics for both sexes combined, as is evident in the graph the distribution is slightly skewed and the confidence limits are quite narrow as indicated by the graph and table 15 above. Due to the slight skewness and narrow confidence limits interpretations must be regarded tenuously. The second histogram represents the distribution of ๐ฝ๐พฬ for women. As is visible in the graph, the distribution of scores is almost normal and the confidence interval is much wider, therefore, inferences regarding the female sample may be made with more rigour as the population estimation deduced from the bootstrap is normally distributed. The histogram regarding male rankings of VA is quite skewed and the confidence interval quite narrow, this may however be due to the reduction in variance caused by the oversampling procedure and thus must be interpreted cautiously.
Figure 15: Histograms representing the distribution of bootstrapped ฮธ_W ฬ VA scores for both sexes combined (left), women (centre), and men (right)
0 1 2 3 4 5 6 7
0 0.2 0.4
Density
Normal(0.143,0.062)
0 0.5 1 1.5 2 2.5 3 3.5
0 0.2 0.4 0.6 0.8
Density
Normal(0.366,0.116)
0 1 2 3 4 5 6 7 8 9
0 0.2 0.4 0.6
Density
Normal(0.108,0.063)
Visual Stimulus Masculinity
A Friedmanโs test indicated no significant difference between the ranked stimulus items when contrasting the sexes (ฯยฒ=6.86, df=5, NS), this was also not significant after bootstrapping was applied (๐๐ฅ2= 6.02, df=5, NS). Therefore, the null hypothesis that there is no significant difference between the ranked items was not rejected, and the conclusion that men and women may rank the items differently was drawn.
As before, the Friedmanโs result is not congruent with the measures of ๐ฬ as agreement between the sexes was indicated to be significant although this agreement was very small and only just significant (see table 16). Subsequently, although men and women combined may have indicated significant agreement, men and women independently, did not show significant concordance before the bootstrap was applied. However, once bootstrapped, ๐๐ฬ
for both sexes independently and both sexes combined indicated significant concordance. The bootstrap increased the degree of agreement amongst all the groups. Men showed greater agreement with regards to VM, sustaining predictions made earlier.
Table 16:Summary of bootstrapped (B=10000) ๐พฬ for rankings of visual masculinity for the sexes
Sex ๐ฬ P ๐๐ฬ P SE Lower
confidence limit
Upper confidence
limit Decision Men and
women
combined 0.072 0.05 0.107 0.001 0.047 0.034 0.213 Reject
Women 0.065 NS 0.108 0.05 0.055 0.026 0.237 Reject
Men 0.163 NS 0.201 0.001 0.089 0.057 0.4 Reject
The histograms indicating the distributions for ๐ฝ๐พฬ for VM for both sexes combined and independent are slightly positively skewed, therefore conclusions must be drawn tentatively regarding all the results. The confidence intervals for both sexes combined and women are slightly narrower than the confidence interval for the male sample, indicating a wider 95%
confidence interval for the estimation of the real world ๐ฝ๐พฬ statistic for men.
Figure 16: Histograms representing the distribution of bootstrapped ฮธ_W ฬ VM scores for both sexes combined (left), women (centre), and men (right)
Scent Stimulus Attractiveness
In comparing the sexes for a difference in the rankings of stimulus items a Friedmanโs test found no significant difference both before the bootstrap (ฯยฒ= 0.21, df=5, NS) and after (๐๐ฅ2=3.4, df=5, NS). Thus, indicating that the null hypothesis should fail to be rejected and that there is no significant difference between the SA ranked items and thus, possibly a difference between the sexes.
This result is supported by the ๐พฬ statistics reported in table 17. Per these results, men and women do not rank the attractiveness of scent concordantly and this result is sustained after the application of bootstrapping. Furthermore, before the bootstrap, both men and women independently were not concordant in their rankings of scent attraction, however, once bootstrapped, significant agreement was indicated, although the degree of agreement for both men and women was small.
Table 17: Summary of bootstrapped (B=10000) ๐พฬ for rankings of scent attractiveness for the sexes
Sex ๐ฬ P ๐๐ฬ P SE Lower
confidence limit
Upper confidence
limit Decision Men and
women
combined 0.019 NS 0.029 NS 0.020 0.005 0.081 fail to
reject
Women 0.086 NS 0.129 0.02 0.056 0.038 0.254 Reject
Men 0.061 NS 0.106 0.05 0.066 0.022 0.280 Reject
0 1 2 3 4 5 6 7 8 9
0 0.1 0.2 0.3
Density
Normal(0.107,0.047)
0 1 2 3 4 5 6 7 8
0 0.2 0.4
Density
Normal(0.108,0.055)
0 1 2 3 4 5
0 0.2 0.4 0.6
Density
Normal(0.202,0.089)
The histograms in figure 17 below, represent the distribution of ๐๐ฬ for women and men. For women, the distribution is positively skewed, although, much less skewed than the distribution for men. The confidence interval however is somewhat narrower than the confidence interval for men, as indicated in table 18. For men, the distribution is quite skewed with most of the distribution falling at the lower end of the confidence interval. These resultant distributions estimated to represent the population by the bootstrap are not normally distributed indicating that the results should be compared tenuously given the assumption that the population is normally distributed.
Figure 17: Histograms representing the distribution of bootstrapped ฮธ_W ฬ SA scores women (left), and men (right)
Scent Stimulus Masculinity
The Friedmanโs test for the masculinity of scent between stimulus items for the sexes indicated that both before bootstrapping (ฯยฒ=4.71, df=5, NS) and after bootstrapping (๐๐ฅ2=5.133, df=5, NS) the distribution of stimulus items were not significantly different, therefore a conclusion can be tenuously drawn that men and women ranked the masculinity of the scent differently.
According to the summary of ๐ฬ statistics for the agreement between the sexes regarding the masculinity of scent, there was no significant agreement amongst sexes combined, men or women. Furthermore, the only original outcome that was not sustained by the bootstrap was concordance amongst men and women combined which increased with the bootstrap and is only just significant with a very small degree of agreement.
0 1 2 3 4 5 6 7 8
0 0.1 0.2 0.3 0.4
Density
Normal(0.129,0.056)
0 1 2 3 4 5 6 7 8
0 0.2 0.4 0.6
Density
Normal(0.106,0.066)
The distribution of ๐๐ฬ regarding masculinity rankings for both sexes combined is illustrated in figure 18 adjacent, indicates a slight positively skewed distribution. This indicates that most estimated ๐๐ฬ statistics lie below the mean and closer to the lower confidence limit. A skewed distribution violates the assumption of normality and therefore any conclusions drawn cannot be substantially indicative of the real-world population. Furthermore, the confidence interval is quite narrow providing a small
window for which to be 95% certain of a real-world estimate.