top of page
This site was designed with the
.com
website builder. Create your website today.Start Now

Hypothesis Testing

The first social network feature that we consider about is the size of the relationship graph. As Figure 3.1 shows, we define the relationship graph of two users by combining the users’ ego network graphs. Tong et al. has showed that in a proper interval number, more friends (social interactions) lead to more sociometric popularity in the online social world. From the view of psychologists, Campfield found that people involved in cyber bullying (cyber bullies, victims, and bully/victims) have lower positive social interaction. We would expect the size of the relationship graph of the cyber bullying group to be smaller than that of the non-bullying group. Therefore, we hypothesize:

 

H.1 Compared to the non-bullying group, the relationship graphs of the cyber bullying group (a) will have less nodes (smaller size); and (b) will have less number of edges (less connections).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hypothesis 1 

 

The second social network feature that we use to compare the cyber bullying and the non-bullying group is the relationship strength. Prior research have reported that users are often cyber bullied by someone whom they know in real life but not considered as close friends. While it has been shown that users are more likely to know each other in real life if they have higher tie strength in online social networks, it has also been found that ‘close friends’ exchange lesser number of messages via online networks. Hence, we expect:

 

H.2 In the relationship graphs, (a) cyber bullying group might have a higher online relationship tie score than the non-bullying group (know each other in real world); also (b) cyber bullying group might have more exchanged messages (not close friend) than the non-bullying group.

Hypothesis 2

The third social network feature that we take into account is the online social activity of users. In the early 1980s, Olweus suggested that externalized behaviors such as argumentativeness and disruptiveness as well as internalized symptoms (e.g loneliness, low self-esteem, and emotional problems) may invite or reinforce peer victimization. Campfield found that cyber victims have emotional and behavioral difficulties as well as loneliness, which may be similar to the vulnerability factors that make face-to-face youth easy targets for cyber bullies. Furthermore, loneliness has been associated with increased online activity, for example, lonely individuals were more likely to report making online friends. It is also known that a user with a higher out-degree centrality posts to more users (higher social activity) than those who have lower out degree centrality. Therefore, we hypothesize that:

 

H.3 The social activity (out-degree centrality) of cyber victims is higher than that of those in other roles (cyber bullies, non-bullying senders, and

non-bullying receivers).

Hypothesis 3

Table 6: Overall results for hypothesis testing

Table 5: Chi-square test for H.3

Table 4: Results for H.2(b)

Table 2: Results for H.1(b)

Table 3: Results for H.2(a)

Table 1: Results for H.1(a)

We found the whole results of hypothesis testing (the general outcomes are shown in Table 6) open a door to understand cyber bullying with social network features which we can explain with the related social science literature. The data from H.1(a) shows that the users in the cyber bullying group have low social interaction. According to Campfield's research, people who are involved in cyber bullying (cyber bully, victim, and bully/victim) have lower global self-esteem and overall sense of self-worth than the non-bullying group, and their positive social interactions (online and offline) are limited. In H.2(a), we found that users in cyber bullying group have higher proportion of common online friends than non-bullying group, which could be seen as these users have higher possibility to know each other offline than non-bullying group. This result aligns with the findings of previous works which state that people involved in cyber bullying typically know each other in real life. The great significance in the number of exchanged posts also has been found for H.2(b). The cyber bullying group contains more exchanged posts than non-bullying group, this result indicates that users in cyber bullying group have less closer relationship in real world. The Chi-square test result for H.3 (P = 0.000169) proves that in the cyber bullying conversations, victims often have more activity than bullies. The reasons for this result could be found in the psychological literature. According to Lopez and DuBois [27], the lonely individuals are likely to have more online activity for emotional support and heightened satisfaction. In conclusion, a user who has a higher out-degree centrality in the relationship graph of conversations might be more likely to be bullied.

bottom of page