BEHS ANNO BIB

Computers in Human Behavior 111 (2020) 106439

Available online 25 May 2020 0747-5632/© 2020 Elsevier Ltd. All rights reserved.

Profiles of adolescent traditional and cyber bullying and victimization: The role of demographic, individual, family, school, and peer factors

Yue Ding a, Dongping Li a, *, Xian Li b, Jiale Xiao a, Haiyan Zhang c, Yanhui Wang d

a School of Psychology, Central China Normal University, Wuhan, Hubei, 430079, China b School of Medicine, Stone Brook University, Stony Brook, NY, 11794, USA c School of Psychology, State University of New York, Cortland, NY, 13045, USA d School of Education Science, Jiaying University, Meizhou, Guangdong, 514015, China

A R T I C L E I N F O

Keywords: Traditional bullying and victimization Cyber bullying and victimization Latent profile analysis Multiple risk exposure

A B S T R A C T

Cyber bullying and victimization are prevalent in the daily lives of a large number of individuals in the current digital age. Both traditional and cyber bullying and victimization among adolescents have raised global concern. Accumulating evidence suggests that there are inter-individual differences between patterns of bullying and victimization among adolescents. However, previous research has primarily relied on variable-centered ap- proaches and failed to reveal the heterogeneity among groups with regard to bullying and victimization. Using a cross-sectional design, we included traditional and cyber bullying and victimization as indicators and employed a person-centered approach to identify distinct subgroups and their associations with demographic, individual, family, school, and peer factors. A total of 1,529 Chinese adolescents (Mage ¼ 14.74 years, SD ¼ 1.48) partici- pated in the study. Latent profile analysis identified three profiles that evidenced heterogeneity of bullying and victimization groups: uninvolved group (92%), high traditional bully-victims group (6%), and high cyber bully-victims group (2%). The findings suggest that these distinct subgroups can be predicted by factors from multiple do- mains. High traditional bully-victims can be predicted by male, younger age, high depressive symptoms, low classmate support, and high deviant peer affiliation. High cyber bully-victims can be predicted by male, younger age, high depressive symptoms, high interparental conflict, and high deviant peer affiliation. In addition, interparental conflict, parental warmth and acceptance, school climate, and peer attachment can further differentiate high traditional bully-victims from high cyber bully-victims. Our results provide implications for tailored prevention and intervention strategies to reduce adolescent bullying and victimization.

1. Introduction

Bullying and victimization have become a global public health concern. Evidence indicates that about 10%–30% or even more children and adolescents worldwide have been affected over the past decades (Cook, Williams, Guerra, Kim, & Sadek, 2010; Kowalski, Giumetti, Schroeder, & Lattanner, 2014; Schoeler, Duncan, Cecil, Ploubidis, & Pingault, 2018). Bullying and victimization bring a series of social adjustment problems (e.g., internalizing and externalizing problems), which can seriously hinder physical and mental health in children and adolescents (Schoeler et al., 2018). Two forms of bullying and victimi- zation have been identified: traditional and cyber. Traditional bullying usually occurs in an offline environment, primarily in schools, and can include direct physical and verbal aggression, as well as indirect

relational aggression (Olweus, 1994). In recent years, cyber bullying and victimization have gradually gained researchers’ attention. Cyber bullying and victimization refer to receiving repeated hostile or aggressive messages from individuals or groups through electronic or digital media (Kokkinos & Antoniadou, 2019). Despite the overlap be- tween cyber and traditional forms of bullying and victimization (Gini, Card, & Pozzoli, 2017), cyber bullying and victimization have unique features. For example, cyber bullies can be anonymous and may be less aware of the consequences of cyber bullying because of a lack of direct feedback for their actions (Kokkinos & Antoniadou, 2019; Kowalski et al., 2014). There have been an increasing number of studies on bullying and victimization in the last two decades; however, most of these used variable-centered approaches, ignoring the heterogeneity among forms of bullying and victimization, and seldom systematically

* Corresponding author. School of Psychology, Central China Normal University, Wuhan, 430079. China. E-mail address: [email protected] (D. Li).

Contents lists available at ScienceDirect

Computers in Human Behavior

journal homepage: http://www.elsevier.com/locate/comphumbeh

https://doi.org/10.1016/j.chb.2020.106439 Received 5 December 2019; Received in revised form 2 May 2020; Accepted 22 May 2020

Computers in Human Behavior 111 (2020) 106439

2

examined the associations of demographic, individual, family, school, and peer factors with different patterns of bullying and victimization. Therefore, the present study employed a person-centered approach to explore the subgroups of bullying and victimization, as well as the unique risk or protective factors for each subgroup.

1.1. Profiles of bullying and victimization

Most previous studies that investigated bullying and victimization used variable-centered approaches. Variable-centered approaches can provide valuable information regarding mean level differences across students who are exposed to bullying and victimization. However, these approaches often fall short in precision and ignore variations in stu- dents’ experiences with bullying and victimization (Moses & Williford, 2017). Compared to variable-centered approaches, person-centered approaches are thought to produce more accurate accounts of bullying and victimization (Nylund, Asparouhov, & Muth�en, 2007). They can capture the variation in students’ experiences, which can better reflect the richness and complexity of adolescents’ real lives. Thus, it is necessary to explore the specific patterns of bullying and victimization with person-centered approaches.

Typical person-centered approaches include cluster analysis and latent class analysis (LCA). In the present study, we employed latent profile analysis (LPA, a variant of LCA) to explore patterns of bullying and victimization. LPA is a mixture modeling strategy that groups in- dividuals into homogeneous profiles based on observed responses to a set of indicators (Kochel, Ladd, Bagwell, & Yabko, 2015). Unlike LCA, which is based on categorical indicators, LPA can provide more valuable information regarding bullying and victimization based on continuous indicators. Additionally, unlike cluster analysis which assigns in- dividuals to clusters absolutely, LPA considers the individual probability of belonging to each profile (i.e., rate of classification uncertainty). Furthermore, LPA does not require researchers to specify the number of profiles in advance, which is more exploratory than other methods. Therefore, researchers can determine the best model fit based on a set of statistical criteria and empirical indicators (Li et al., 2017).

A growing number of studies focused on bullying and victimization have employed person-centered approaches, which advanced the liter- ature in this field (Moses & Williford, 2017). However, there are several limitations of these studies that need to be noted. First, many studies focused only on bullying or victimization, without examining both (e.g., Bradshaw, Waasdorp, & Johnson, 2015; Bradshaw, Waasdorp, & O’Brennan, 2013; Wang, Iannotti, & Luk, 2012). For example, Bradshaw et al. (2015) took LCA to examine patterns of bullying victimization, and identified four classes—low victimization class, multiple victimization class, relational victimization class, and physical victimization class. Although this study explored the overlaps between verbal, relational, physical, and electronic forms of bullying victimization and revealed the heterogeneity among victims, they did not focus on bullies. In fact, there is an overlap between bullying and victimization, which supports our decision to include indicators of both bullying and victimization in an LPA framework (Rodríguez-Hidalgo, Pantale�on, & Calmaestra, 2019). A few studies identified subgroups based on both bullying and victimiza- tion indicators, and most identified four subgroups—uninvolved stu- dents, bullies, victims, and bully-victims (e.g., Kochel et al., 2015; Lovegrove, Henry, & Slater, 2012)—with some exceptions (Ettekal & Ladd, 2017; Giang & Graham, 2008). However, some of these studies conducted LPA only based on total bullying scores and total victimiza- tion scores (Kochel et al., 2015), which ignored the heterogeneity among different forms of bullying and victimization. Thus, it is necessary to break down bullying and victimization by types. Some of these studies conducted LPA based on six single items regarding bullying and victimization (Lovegrove et al., 2012), which might lead to lower measurement sensitivity than using multiple items. Even in the same sample, the prevalence rates of bullying and victimization measured by single items are likely to be lower than multi-item checklists (Kowalski

et al., 2014). Second, prior studies seldom took into account both traditional and

cyber bullying and victimization at the same time. Cyber bullying and victimization have only gained attention in the past few years, yet given that adolescents have more access to cyberspace with the arrival of the digital age, it is essential to include both traditional and cyber forms of bullying and victimization in research. Antoniadou, Kokkinos, and Fanti (2019) identified four subgroups—uninvolved students, bullies, victims, and bully-victims, based on traditional and cyber bullying and victimi- zation, via LPA. Their findings revealed the overlaps between traditional and cyber bullying and victimization. However, they used total scores for traditional bullying and victimization, which ignored the heteroge- neity among different forms of traditional bullying and victimization. Previous studies suggested that cyber bullying and victimization were more strongly correlated to specific forms of traditional bullying and victimization (e.g., relational bullying and victimization—an indirect form; Modecki, Minchin, Harbaugh, Guerra, & Runions, 2014). There- fore, including different forms of traditional bullying and victimization may be an ideal approach to examine the overlaps and differences be- tween cyber bullying and victimization and specific forms of traditional bullying and victimization. In addition, Lovegrove and Cornell (2014) identified four similar subgroups,via LCA. The difference is that the four classes were based on 10 indicators, including overall, physical, verbal, social, and cyber bullying and victimization. Although they included comprehensive forms of bullying and victimization, they only used a single item to measure each form, which might have led to lower prevalence rates (Kowalski et al., 2014). To address the limitations of prior studies, we employed LPA to explore latent profiles of bullying and victimization and used both traditional and cyber forms of bullying and victimization as indicators. Moreover, we used multiple items to mea- sure levels of bullying and victimization, which improved measurement sensitivity.

1.2. Multiple risk and protective factors of bullying and victimization

Studies have indicated that becoming a bully, victim, or bully-victim are not random events, but can be predicted by multiple risk and pro- tective factors (Arseneault, Bowes, & Shakoor, 2010). Therefore, it is of great importance to investigate the predictors of bullying and victimi- zation. A number of variable-centered studies have identified a set of risk and protective factors of bullying and victimization. However, studies also found that distinct subgroups, such as victims and bully-victims, have different risk and protective factors (Arseneault et al., 2010), and person-centered approaches can further detect the nuances of these factors among different subgroups (Moses & Williford, 2017).

Many theories (e.g., frustration-aggression theory, general aggres- sion model) explored the risk and protective factors of bullying and victimization, which were of great significance for understanding the occurrence of bullying and victimization. However, researchers emphasized that bullying and victimization are ecological phenomena and should be considered as consequences of the interplay between in- dividuals and ecological systems, whether in traditional or cyber forms (Espelage, 2014; Espelage & Swearer, 2010). From an ecological perspective, family, school, and peer environments provide social con- texts for adolescents, and thereby shape their perceptions and attitudes toward behaviors like bullying (Erginoz et al., 2013). In fact, a number of research—not only empirical research (e.g., Cross et al., 2015; Hong et al., 2016) but also meta-analytic and narrative reviews (e.g., Cook et al., 2010; Guo, 2016; Hong & Espelage, 2012; Zych, Farrington, & Ttofi, 2018)—applied ecological perspective and confirmed that the risk of being involved in bullying and victimization can be predicted by in- dividual characteristics and social contexts. For example, Cook et al.’s (2010) meta-analytic study found that the predictors of traditional bullying and victimization included both individual (e.g. self-related cognition, externalizing and internalizing problems, social

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

3

competence, academic performance) and contextual factors (e.g. family environment, school climate, and peer status). Guo (2016) conducted a meta-analysis of cyber bullying and victimization, and also found mul- tiple predictors from demographic, individual, family, school, and peer domains. Therefore, it is important to explore the risk and protective factors of adolescent bullying and victimization from an ecological perspective. Of note, there are many overlaps in risk/protective factors of traditional and cyber bullying and victimization. We will describe the specific relations in the following review only when the risk/protective factors differ for traditional and cyber forms.

A number of studies have found associations between demographic, individual, family, school, and peer factors and bullying and victimi- zation. Among demographic factors, boys were more likely to be involved in traditional bullying and victimization than girls, and younger students were more likely to be involved compared to older ones (Arseneault et al., 2010; Bradshaw et al., 2013; Murphy, Laible, & Augustine, 2017). However, the current evidence on whether gender and age can predict cyber bullying and victimization is controversial. Tokunaga’s (2010) review indicated that there is no significant associ- ation of gender and age with cyber bullying or victimization, while other meta-analysis research suggested that boys and older students were more likely to be involved in cyber bullying and victimization than girls and younger students (Barlett & Coyne, 2014; Kowalski et al., 2014). Evidence from meta-analytic and narrative reviews also found that low socioeconomic status was positively associated with traditional and cyber bullying and victimization (Kowalski et al., 2014; Tippett & Wolke, 2014). Additionally, living in a single-parent household is also a risk factor for traditional (not cyber) bullying and victimization (Stalmach, Tabak, & Radiukiewicz, 2014; Yang, Stewart, & Kim, 2013).

Individual characteristics are closely related to experiences of bullying and victimization. As for bullying, adolescents who reported more feelings of anger/frustration, sensation-seeking, and lower levels of effortful control were more likely to be bullies (Lovegrove et al., 2012). Meta-analytic research also indicated that anger was a risk factor for cyber bullying perpetration (Kowalski et al., 2014). Moral disen- gagement could predict both traditional and cyber bullying, and the association was confirmed by meta-analytic research (Gini, Pozzoli, & Hymel, 2014; Killer, Bussey, Hawes, & Hunt, 2019). Moreover, adoles- cents who engaged in traditional and cyber bullying more frequently often engaged in more other delinquent behaviors as well (Barker, Arseneault, Brendgen, Fontaine, & Maughan, 2008; Kowalski et al., 2014). As for victimization, researchers indicated that students who had lower levels of self-esteem and worse social skills were more likely to be bullied, both traditional and cyber (Bilsky et al., 2013; Brewer & Ker- slake, 2015; Low & Espelage, 2013; Reijntjes, Kamphuis, Prinzie, & Telch, 2010). As for bullying and victimization, evidence from meta-analytic research supported the role of internalizing problems in predicting traditional and cyber bullying and victimization, and the effect sizes were slightly greater in cyber bullying and victimization than traditional forms (Cook et al., 2010; Guo, 2016). Moreover, several meta-analytic studies explored the role of academic performance, and the results indicated that good academic performance was related to low traditional and cyber bullying, low traditional victimization, but not cyber victimization (Cook et al., 2010; Kowalski et al., 2014; Nakamoto & Schwartz, 2010; Zych et al., 2018). Additionally, Internet use is a special factor for cyber bullying and victimization. Individuals who spent more time on the Internet were more likely to be involved in cyber bullying and victimization (Bauman, 2009; Chen, Ho, & Lwin, 2016).

Family, as the most proximal and influential context for individual development (Bronfenbrenner, 1986), is a significant predictor for bullying and victimization (Arseneault et al., 2010). Meta-analytic research supported the impact of family environment on traditional and cyber bullying and victimization, especially positive parenting styles which were protective against bullying and victimization (Cook et al., 2010; Guo, 2016; Lereya, Samara, & Wolke, 2013; Zych et al., 2018). Adolescents raised by parents with low warmth/acceptance, low

behavioral control, and high psychological control were more likely to be involved in both traditional and cyber bullying and victimization (G�omez-Ortiz, Rey, Casas, & Ortega-Ruiz, 2014; Kokkinos, Antoniadou, Asdre, & Voulgaridou, 2016). Additionally, adolescents who reported insecure parent-child attachments had higher risks of being involved in traditional and cyber bullying and victimization (Earl & Burns, 2009; Hemphill, Tollit, Kotevski, & Heerde, 2014; Murphy et al., 2017; Ste- vens, Bourdeaudhuij, & Oost, 2002). Furthermore, interparental conflict has been positively associated with traditional and cyber bullying (Hemphill et al., 2014; Kretschmer et al., 2015; Stevens et al., 2002), as adolescents often learn how to interact with others from their parents.

Adolescents spend a large proportion of their time at school, second only to family environments, and the majority of bullying and victimi- zation occurs in school (Arseneault et al., 2010). Studies found that students’ reports of bullying and victimization were influenced by their perceived school climate. The more positive the perception of one’s school climate, the lower the risk of being involved in bullying and victimization incidents, both traditional and cyber (Bradshaw, Sawyer, & O’Brennan, 2009; Holfeld & Leadbeater, 2017; Klein, Cornell, & Konold, 2012). Evidence from meta-analytic research supported this association, and the effect sizes were slightly greater in traditional bullying and victimization than cyber forms (Zych et al., 2018). Studies indicated that students who felt more connected to their schools were less likely to be involved in both traditional and cyber bullying and victimization (Hong & Espelage, 2012; Thapa, Cohen, Guffrey, & Hig- gins-D’Alessandro, 2013; Kowalski et al., 2014). Moreover, high levels of support from teachers and classmates have been shown to decrease the risk of traditional and cyber bullying and victimization (Flaspohler, Elfstrom, Vanderzee, Sink, & Birchmeier, 2009; Kowalski et al., 2014; Yubero, Ovejero, & Larranaga, 2010).

During adolescence, the significance of peer influence increases, as individuals gradually seek independence from their parents. Thus, peers are an important source of social influence associated with bullying and victimization (Wang, Iannotti, & Nansel, 2009). The impact of peer factors on traditional and cyber bullying and victimization was also confirmed by meta-analytic research, and the effect sizes were greater in traditional forms (Cook et al., 2010; Guo, 2016). According to the friendship protection hypothesis, adolescents with friends are less likely to be involved in bullying and victimization than those without friends (Hodges, Boivin, Vitaro, & Bukowski, 1999). Similarly, previous studies found that secure peer attachments decreased the risk of being involved in both traditional and cyber bullying victimization (Murphy et al., 2017; Wright, Kamble, & Soudi, 2015). Additionally, studies suggested that adolescents who were affiliated with more deviant peer groups were more likely to be involved in traditional bullying and victimization (Hong, Kim, & Piquero, 2017); however, this association has not been examined in the context of cyber bullying and victimization.

Based on the theories and literature reviewed above, closer exami- nation of multiple predictors for bullying and victimization is necessary. Furthermore, multiple risk/protective factors often co-occur with each other, and if we only focus on a singular risk/protective factor and ignore others, the impact of that risk/protective factor may be exag- gerated. Moreover, an intervention for a single risk/protective factor is less effective than focusing on multiple risk/protective factors simulta- neously (Evans, Li, & Whipple, 2013). Even though several studies considered the predictors of bullying and victimization subgroups (e.g., Lawson, Alameda-Lawson, Downer, & Anderson, 2013; Lovegrove et al., 2012), the factors included were not comprehensive. To address this limitation, we included 25 typical risk and protective factors from 5 domains to investigate their prediction of profile memberships. For how the risk and protective factors were selected, please refer to the Method section.

1.3. The present study

Given the heterogeneity among forms of bullying and victimization,

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

4

we employed a person-centered approach to identify subgroups among adolescents. Specifically, this study had two purposes. The first was to identify subgroups of bullying and victimization. We modeled six in- dicators (i.e., traditional direct bullying, traditional indirect bullying, cyber bullying, traditional direct victimization, traditional indirect victimization, and cyber victimization) as observed variables. Based on previous studies (e.g., Kochel et al., 2015; Lovegrove & Cornell, 2014; Lovegrove et al., 2012), we hypothesized the following four profiles: uninvolved profile, bullys profile, victims profile, and bully-victims profile.

The second purpose was to identify the predictors for different pro- files. Such predictors can provide useful information for identifying distinct groups and informing intervention design. We included 25 risk and protective factors of bullying and victimization from 5 domains (i.e., demographic, individual, family, school, and peer). For the diversity and complexity of risk and protective factors, we did not make a hypothesis of their results. Furthermore, a number of studies suggested that stu- dents’ risk levels for bullying and victimization peak in adolescence (Bradshaw et al., 2013; Lawson et al., 2013); thus, adolescence is a key stage for prevention of and intervention for bullying and victimization. Therefore, we used adolescents as research participants to explore the status and predictors of bullying and victimization.

2. Method

2.1. Participants

Data for this study were collected as part of the Adolescent Bullying and Victimization in the Digital Age project. Participants were 1,529 students (52% male) in grades 7 through 11 from seven middle and high schools in Wuhan, China. Two classes in each grade at each school were randomly selected to take part in the present study. The mean age of the participants was 14.74 years (SD ¼ 1.48, range ¼ 12–19). In terms of family structure, 88.9% of the students came from intact families. Additionally, 45.1% of the students’ fathers and 54.7% of their mothers had less than a high school education; 45.3% of their fathers and 47.1% of their mothers had an unskilled or semi-skilled occupation. Over 95% of the students agreed to participate in this study.

2.2. Procedure

This study was approved by the Research Ethics Committee of the corresponding author’s institution. Verbal consent was obtained from school administrators, teachers, and students before data collection. Because this study was part of the regular psychological health educa- tion courses in schools, teachers had consulted with parents in advance. Students answered printed questionnaires independently with the guidance of well-trained research assistants during a single class period. Students were informed that their participation was completely volun- tary and they could decline participation at any time. Given that it is a challenge to conduct studies with multiple risk factors, we followed Li et al.’s (2017) suggestions to balance the breadth and depth of the questionnaire with the response quality of the participants. Specifically, we selected measures that were relatively short but still retained important concepts to assess risk factors, which could reduce response load. We also used various respondent-friendly designs, such as using simple but clear language and varying the pace and type of questions. Additionally, research assistants provided necessary encouragement for participants during the process. School counselors were available for potentially distressed students. Finally, we provided small incentives for all participants, regardless of survey completion.

2.3. Measures

Traditional bullying and victimization. To estimate the preva- lence of traditional forms of bullying and victimization, we used multi- item scales that asked the frequency of specific behaviors representing

bullying and victimization, which were more reliable and valid than single item scales (Thomas, Connor, & Scott, 2015). Scales adapted from the Revised Olweus Bully/Victim Questionnaire were used to assess adolescents’ experiences of traditional bullying and victimization (Olweus, 1996). Students were asked to rate how often the described events had occurred at school during the past 6 months using a 5-point scale, ranging from 1 ¼ never to 5 ¼ six times or more. Responses to the items were averaged, with higher scores indicating higher frequency of traditional bullying or victimization. Traditional direct bullying was assessed using 3 items (α ¼ 0.78; e.g., “I hit, kicked, or pushed some- body”). Traditional indirect bullying was assessed using 6 items (α ¼ 0.83; e.g., “I spread false rumors about somebody behind his/her back”). Traditional direct victimization was assessed using 3 items (α ¼ 0.83; e. g., “Some students hit, kicked, or pushed me around”). Traditional in- direct victimization was assessed using 5 items (α ¼ 0.83; e.g., “Some students spoke ill of me behind my back”). This questionnaire has demonstrated validity in assessing traditional bullying and victimization in previous research (Li et al., 2018; Solberg & Olweus, 2003).

Cyber bullying and victimization. To avoid the limitations asso- ciated with single item measures, we followed the suggestion of Kowalski et al.’s (2014) systematic review. Specifically, we used multi-item behavioral checklists that shared a response scale (e.g., 1 ¼ never to 5 ¼ several times per week) and utilized the same reporting time frame (e.g., during the past 6 months) to measure cyber bullying and victimization. This study used scales adapted from the Cyber Bullying Questionnaire to assess adolescents’ experience of cyber bullying and victimization (Menesini, Nocentini, & Calussi, 2011). Students were instructed to rate how often they had experienced the described events during the past 6 months using a 5-point scale, ranging from 1 ¼ never to 5 ¼ six times or more. Responses across the items were averaged, with higher values representing more frequent cyber bullying or victimiza- tion. Within the questionnaire, cyber bullying was assessed using 6 items (α ¼ 0.80; e.g., “cursing at somebody on social networking site such as blog”). Cyber victimization was assessed using 6 items (α ¼ 0.81; e.g., “receiving hurtful text messages or receiving threats through e-mail”). A previous study demonstrated that this questionnaire has good reliability and validity (Menesini et al., 2011).

Multiple risk and protective factors. Multiple methods were used to identify the risk and protective factors. First, we performed literature searches of multiple databases (e.g., PsycINFO, Web of Science, and CNKI), and the search terms included (bully* or victim*) AND (adolescen*). Second, we identified risk and protective factors by reading both empirical articles and literature reviews published in Chinese and English languages. To ensure the feasibility of this study, we did not exhaust all possible factors. We included and excluded risk/ protective factors meeting the following criteria: (a) the factors should belong to demographic, individual, family, school, or peer domains; (b) they should be consistent with the developmental stage of adolescents; (c) they must be closely related to traditional and/or cyber bullying victimization (i.e., factors with very small or non-significant effects were excluded); and (d) they should be typical and unique (i.e., factors seldom be investigated in the literature or belong to the superordinate concept of the selected factors were excluded). Third, we used a pilot study to examine the correlations between potential factors and different forms of bullying and victimization. Only factors significantly related to at least one form of bullying and victimization were included in the formal study. The final 25 factors have covered most variables examined in previous research and might be useful in predicting distinct profiles. Specifically, we included the following risk and protective factors from five domains: demographic (i.e., gender, age, family structure, and socioeconomic status), individual (i.e., Internet use, effortful control, sensation-seeking, anger/frustration, social skills, moral disengagement, self-esteem, depressive symptoms, delinquency, and academic performance), family (i.e., interparental conflict, warmth and acceptance, behavioral control, psychological control, and parent- child attachment), school (i.e., school climate, school connectedness,

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

5

teacher support, and classmate support), and peer factors (i.e., peer attachment and deviant peer affiliation). Among demographic factors, gender was measured as a dichotomous variable (0 ¼ female and 1 ¼ male), as was family structure (0 ¼ non-intact family and 1 ¼ intact family). For socioeconomic status, a factor score was derived using principal component analysis of fathers’ education, mothers’ education, father’s occupation, mother’s occupation, and family income. Higher scores indicated higher family socioeconomic status. For simplicity, all other risk and protective factors and how they were measured are listed in Table 1. The reliability and validity of all the instruments have been validated in samples of Chinese adolescents.

2.4. Analytic strategies

We conducted preliminary analyses in SPSS 17.0. In the present study, less than 1% of data were missing, which were handled with mean imputation (Little & Rubin, 2002). Differences in participants by school were also tested. Then, we identified profiles and their predictors step by step. In the first step, we conducted LPA in Mplus 7.4 (Muth�en & Muth�en, 1998-2015) to identify subgroups of adolescents with similar patterns across the six bullying and victimization indicators: traditional direct bullying, traditional indirect bullying, cyber bullying, traditional direct victimization, traditional indirect victimization, and cyber victimization. The scores of the six indicators were standardized for ease of interpretation. LPA estimates two sets of parameters (i.e., latent profile proportions, and conditional means) to break down profiles. The latent profile proportions refer to the percentage of the sample expected to be members of specific latent profiles, and the conditional means represent the mean values of profile members on each indicator (Flah- erty & Kiff, 2012). A two-profile model was fit first, followed by three, four, five, etc. We added additional profiles until there was no empirical or conceptual improvement, such as the proportion of profiles became too small or the profiles were theoretically meaningless (Lawson et al., 2013). Model solutions for selecting the appropriate number of profiles were evaluated based on a comparison using the following statistical criteria (Nylund et al., 2007): Akaike information criterion (AIC), Bayesian information criterion (BIC), sample-size adjusted Bayesian information criterion (a-BIC), Lo–Mendell–Rubin likelihood ratio test (LRT), and entropy. Models with smaller AIC, BIC, and a-BIC values suggest better fitting solutions. LRT compares a model with k profiles to a model with k–1 profiles. Significant p-values indicate that a model with k profiles has better fit than a model with k–1 profiles. Entropy values represent the precision and classification quality of models. Values range from 0 to 1 and higher values indicate greater model precision and classification. According to Jung and Wickrama (2008), determining the number of profiles depends on many factors and not only on statistical criteria but also on the research question, parsimony, theoretical justification, and interpretability. In the second step, 25 covariates were introduced to the LPA model. To explore which vari- ables could significantly predict profile membership, the model utilized multivariate logistic regression for examining all covariates simulta- neously. Although these variables were specified as predictors, the causal relationships could not be asserted with the present cross-sectional design. Furthermore, to assess the extent to which the type I error rate might challenge our findings, we used false discovery rate control to adjust p-values of multiple risk/protective factors. This approach is more powerful than methods like Bonferroni procedure. (Glickman, Rao, & Schultz, 2014).

3. Results

3.1. Preliminary analyses

Analyses of variance were conducted to explore the differences in participants by school. The differences among schools on traditional and cyber bullying and victimization were significant, F(6, 1522) ¼

Table 1 Summary descriptions of measures by risk and protective factors.

Risk factors Measure (Number of items)

Sample item (Scale) α

Individual Internet use Internet usage time (1) How long do you use

Internet every week in recent six months? (1 ¼ never, 6 ¼ 15 h above)

–

Effortful control

Adopted from Early Adolescent Temperament Questionnaire-Revised ( Bao, Li, Zhang, & Wang, 2015) (16)

It’s hard for me to concentrate on one thing. (1 ¼ strongly disagree, 6 ¼ strongly agree)

.84

Sensation- seeking

Sensation-Seeking Scale ( Li, Zhang, Li, Zhen, & Wang, 2010) (5)

I like rock climbing, bungee and other adventure activities. (1 ¼ strongly disagree, 6 ¼ strongly agree)

.79

Anger/ frustration

Adopted from Early Adolescent Temperament Questionnaire-Revised (EATQ-R; Ellis & Rothbart, 2001) (8)

I lose my temper easily. (1 ¼ strongly disagree, 6 ¼ strongly agree)

.72

Social skills Social skills Scale of Self- Perception Profile for Adolescents (SPPA; Harter, 1988) (6)

It’s hard/easy to make friends. (1 ¼ very hard, 8 ¼ very easy)

.82

Moral disengagement

Moral Disengagement Scale (Bandura, Barbaranelli, Caprara, & Pastorelli, 1996) (32)

It’s reasonable to fight to protect friends. (1 ¼ strongly disagree, 5 ¼ strongly agree)

.88

Self-esteem Rosenberg Self-Esteem Scale (Rosenberg, 1965) (10)

I think I am a valuable person, at least on the same level as others. (1 ¼ strongly disagree, 4 ¼ strongly agree)

.86

Depressive symptoms

Adapted from Children’s Depression Inventory ( Kovacs, 1992; Li et al., 2015) (10)

I have been unhappy in recent two weeks. (1 ¼ seldom, 3 ¼ always)

.78

Delinquency Problem Behavior Scale ( Bao et al., 2015) (12)

Robbing, blackmailing, or threatening others in recent six months. (1 ¼ never, 5 ¼ 5 times and above)

.66

Academic performance

Academic Performance Scale (3)

How about your present grade in mathematics? (1 ¼ extremely bad, 5 ¼ extremely good)

–

Family Interparental conflict

Children’s Perception of Interparental Conflict Scale (CPIC; Grych, Seid, & Fincham, 1992; Chi & Xin, 2003) (15)

I often see my parents quarreling. (1 ¼ strongly disagree, 4 ¼ strongly agree)

.90

Warmth and acceptance

Adapted from Li, Li, Wang, and Bao (2016) (9)

My parents will help or support me when I am in trouble. (1 ¼ strongly disagree, 5 ¼ strongly agree)

.86

Behavioral control

Parental Control Questionnaire (Wang, Pomerantz, & Chen, 2007) (13)

Do your parents often ask you how much time you spend on studying? (1 ¼ never, 5 ¼ always)

.86

Psychological control

Parental Control Questionnaire (Wang et al., 2007) (16)

My parents tell me that I should feel ashamed when I do not behave as they expect. (1 ¼ strongly disagree, 5 ¼ strongly agree)

.86

Parent-child attachment

Chinese version of the Inventory of Parent and Peer Attachment-Short Version (Li et al., 2015) (13)

My parents respect my feelings. (1 ¼ strongly disagree, 5 ¼ strongly agree)

.90

School

(continued on next page)

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

6

2.20–14.57, ps < .05. However, according to Cohen’s (1988) definition, the effect sizes were relatively small (Mpartial ƞ2 ¼ 0.02). In addition, given that the base rate of cyber bullying and victimization was not very high, we did not identify profiles separately by school in subsequent analyses.

3.2. Profiles of bullying and victimization

To explore latent patterns of adolescents’ bullying and victimization, we fit a series of models with two to six profiles that were based on traditional direct bullying, traditional indirect bullying, cyber bullying, traditional direct victimization, traditional indirect victimization, and cyber victimization. All fit indices (i.e., AIC, BIC, a-BIC, entropy, LRT, and profile size) are presented in Table 2. After reviewing the latent profile models specifying two to six profiles, we selected the three- profile model. Specifically, the AIC, BIC, and a-BIC kept declining from two- to six-profile solutions, so these indices could not provide much information regarding the selection. The entropy values were high for all solutions; however, there was a greater increase in the three- profile solution and no apparent improvement after it. The p-value for LRT never became significant; thus, it was not considered when deter- mining the final model. It is not uncommon that statistical indices do not provide clear information on the selection of profiles. To address this situation, Wang and Bi (2018) suggested that one should consider other criteria, such as profile size and interpretability of each solution. For example, they suggested that even if statistical criteria supported the model with k profiles, if the profile size or interpretability was limited,

we should consider the model with k� 1 profiles. Since the minimal profile proportions of the five- and six-profile solutions were smaller than 1%, we did not select these two solutions. Comparing three- and four-profile solutions, the four-profile solution divided the traditional bully-victims into two smaller profiles, while the three-profile solution also identified the traditional bully-victims. Additionally, the two remaining profiles in three- and four-profile solutions were closely resembled in forms and estimated membership rate. Therefore, consid- ering the parsimony and interpretability of the model, the three-profile solution was more reasonable.

Because the probabilities of participants being assigned to a single latent profile identified by LPA were high (>.93), we assigned each participant to the profile associated with the highest posteriori proba- bility (Goodman, 2007). Thus, we could get specific information about the breakdown point by estimating mean value of each profile members on each indicator. The means of the six indicators for the three-profile solution are depicted in Fig. 1. We ran analyses of variance to deter- mine if the three profiles differed in their ratings of each indicator, and the parameter estimates and within-profile indicator means are pre- sented in Table 3. Profile 1, which contained 92% of the sample (n ¼ 1401), was labeled as the uninvolved group. The members in this group had low or even no bullying and victimization. Profile 2, which con- tained 6% of the sample (n ¼ 90), was labeled as the high traditional bully-victims group. The members in this group seldom bullied others online and had low cyber victimization; however, they often bullied or were bullied offline, especially through traditional direct ways. Profile 3, which contained 2% of the sample (n ¼ 38), was labeled as the high cyber bully-victims group. The members in this group usually carried out cyber bullying and suffered cyber victimization as well. At the same time, they also experienced moderate levels of traditional bullying and victimization.

3.3. Predictors of bullying and victimization profiles

We used multivariate logistic regression to explore the predictors for the 25 variables, including 4 demographics factors, 10 individual fac- tors, 5 family factors, 4 school factors, and 2 peer factors. The results of the logistic regression are presented in Table 4.

Table 1 (continued )

Risk factors Measure (Number of items)

Sample item (Scale) α

School climate School Climate Scale (Bao et al., 2015) (25)

Students respect each other. (1 ¼ strongly disagree, 6 ¼ strongly agree)

.88

School connectedness

School Connectedness Scale (Li et al., 2013) (6)

I like this school. (1 ¼ strongly disagree, 6 ¼ strongly agree)

.78

Teacher support

Perceived Social Support Scale (Zimet, Dahlem, Zimet, & Farley, 1988) (4)

Teachers would help me when I encounter problems. (1 ¼ very strongly disagree, 7 ¼ very strongly agree)

.90

Classmate support

Perceived Social Support Scale (Zimet et al., 1988) (4)

Schoolmates can share happiness and sadness with me. (1 ¼ very strongly disagree, 7 ¼ very strongly agree)

.88

Peer Peer attachment

Simple version of the Inventory of Parent and Peer Attachment (IPPA; Raja, McGee, & Stanton, 1992) (13)

My friends respect my feelings. (1 ¼ strongly disagree, 5 ¼ strongly agree)

.87

Deviant peer affiliation

Deviant Peer Affiliation Scale (Li et al., 2013) (8)

How many of your friends smoked in recent twelve months? (1 ¼ none, 5 ¼ all)

.79

Table 2 Fit statistics for latent profile analysis.

Number of profiles Log likelihood Free parameters AIC BIC a-BIC Entropy LRT Proportions Min

2 � 12097.53 19 24233.06 24334.38 24274.02 .978 .123 .08 3 � 11505.16 26 23062.32 23200.96 23118.36 .990 .682 .02 4 � 10995.76 33 22057.52 22233.49 22128.66 .990 .722 .01 5 � 10514.99 40 21109.98 21323.27 21196.20 .992 .182 .00 6 � 10072.14 47 20238.29 20488.91 20339.60 .986 .560 .00

Note. AIC ¼ Akaike information criterion; BIC ¼ Bayesian information criterion; a-BIC ¼ sample-size adjusted Bayesian information criterion; LRT ¼ Lo–Mendell–Rubin adjusted likelihood ratio test.

Fig. 1. Latent profile indicator means for the three-profile solution. Note. TDB ¼ Traditional direct bullying; TIB ¼ Traditional indirect bullying; CB ¼ Cyber bullying; TDV ¼ Traditional direct victimization; TIV ¼ Traditional indirect victimization; CV ¼ Cyber victimization.

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

7

Profile 2 versus Profile 1. Boys were more likely to be high tradi- tional bully-victims than be uninvolved (t ¼ 4.89, OR ¼ 4.66, p < .001). Older students were less likely to be traditional bully-victims (t ¼� 4.19, OR ¼ 0.68, p < .001). High levels of depressive symptoms and deviant peer affiliations increased the risks of being high traditional bully- victims, while high levels of classmate support decreased the risks (t ¼ 2.76, OR ¼ 3.96, p ¼ .006; t ¼ 2.65, OR ¼ 1.96, p ¼ .008; t ¼ � 2.55, OR ¼ 0.75, p ¼ .011). Other factors were not significantly associated with higher or lower risks of being high traditional bully-victims, compared to being uninvolved. After adjusting p-values, the associa- tion between classmate support and profiles became non-significant, but the other significant factors remained significant.

Profile 3 versus Profile 1. Similarly, boys were more likely to be high cyber bully-victims than be uninvolved (t ¼ 3.09, OR ¼ 4.14, p ¼ .002). Older students were less likely to be cyber bully-victims (t ¼ � 2.85, OR ¼ 0.66, p ¼ .004). Higher levels of depressive symptoms, interparental conflict, and deviant peer affiliations increased the risks of being high cyber bully-victims, compared to being uninvolved (t ¼ 2.81, OR ¼ 6.63, p ¼ .005; t ¼ 2.74, OR ¼ 2.50, p ¼ .006; t ¼ 3.41, OR ¼ 3.31, p ¼ .001). Other factors were not significantly associated with higher or lower risks of being high cyber bully-victims, compared to being unin- volved. After adjusting p-values, the significance of the factors did not change.

Profile 3 versus Profile 2. Individuals who experienced higher levels of interparental conflict or warm parenting were more likely to be high cyber bully-victims than high traditional bully-victims (t ¼ 2.41, OR ¼ 2.45, p ¼ .016; t ¼ 2.10, OR ¼ 2.51, p ¼ .036). Positive school climates and secure peer attachments decreased the risks of being a high cyber bully-victim (t ¼ � 2.00, OR ¼ 0.52, p ¼ .046; t ¼ � 2.40, OR ¼ 0.41, p ¼ .016). Other factors were not significantly associated with higher or lower risks of being high traditional bully-victims, compared to high cyber bully-victims. After adjusting p-values, these factors could not significantly distinguish Profile 3 from Profile 2.

4. Discussion

The present study employed a person-centered approach to explore profiles of bullying and victimization and the associations between multiple risk/protective factors and profile memberships. We included both traditional and cyber bullying and victimization indicators, and identified three distinct subgroups: uninvolved group, high traditional bully-victims group, and high cyber bully-victims group. We also found

Table 3 Means and standard deviations of bullying and victimization indicators by latent profile membership.

Uninvolved (n ¼ 1401)

High traditional bully-victims (n ¼ 90)

High cyber bully- victims (n ¼ 38)

F

Traditional direct bullying

1.10 (0.29)a 1.96 (1.12)b 1.47 (0.92)c

192.96***

Traditional indirect bullying

1.06 (0.19)a 1.20 (0.49)b 1.62 (0.92)c

95.43***

Cyber bullying 1.02 (0.09)a 1.06 (0.12)a 1.75 (0.88)b

370.28***

Traditional direct victimization

1.23 (0.40)a 3.86 (0.78)b 2.09 (1.11)c

1433.30***

Traditional indirect victimization

1.17 (0.34)a 1.94 (1.06)b 2.03 (0.94)b

195.58***

Cyber victimization

1.08 (0.19)a 1.19 (0.33)b 2.47 (0.32)c

652.82***

Note. Standard deviations are in parentheses. Means in the same row that share superscript are not significantly different at the p < .05 level. ***p < .001.

Table 4 Results of LPA with covariates—multivariate logistic regressions.

Logit SE t OR p p- adjusted

Profile 2 vs. Profile 1 Demographics

Gender 1.54 0.32 4.89 4.66 <.001 <.001 Age � 0.39 0.09 � 4.19 0.68 <.001 <.001 Family structure 0.38 0.45 0.84 1.46 .402 .628 SES � 0.01 0.15 � 0.08 1.00 .938 .990

Individual Internet use � 0.01 0.10 � 0.05 0.99 .960 .990 Effortful control � 0.25 0.23 � 1.12 0.78 .264 .628 Sensation-seeking � 0.03 0.11 � 0.25 0.97 .799 .951 Anger/frustration 0.23 0.16 1.51 1.26 .130 .542 Social skills � 0.05 0.10 � 0.51 0.95 .611 .764 Moral disengagement

0.26 0.26 1.00 1.30 .318 .628

Self-esteem 0.31 0.31 0.99 1.37 .320 .628 Depressive symptoms

1.38 0.50 2.76 3.96 .006 .050

Delinquency 1.47 1.09 1.34 4.34 .179 .559 Academic performance

� 0.16 0.16 � 0.98 0.86 .328 .628

Family Interparental conflict

0.02 0.22 0.10 1.02 .924 .990

Warmth and acceptance

� 0.22 0.24 � 0.91 0.80 .364 .628

Behavioral control 0.15 0.20 0.74 1.16 .457 .635 Psychological control

0.12 0.22 0.55 1.13 .582 .764

Parent-child attachment

0.26 0.28 0.90 1.29 .366 .628

School School climate 0.20 0.23 0.88 1.23 .378 .628 School connectedness

0.001 0.12 0.01 1.00 .990 .990

Teacher support 0.09 0.12 0.75 1.09 .456 .635 Classmate support � 0.29 0.11 � 2.55 0.75 .011 .055

Peer Peer attachment 0.29 0.20 1.42 1.34 .155 .554 Deviant peer affiliation

0.67 0.25 2.65 1.96 .008 .050

Profile 3 vs. Profile 1 Demographics

Gender 1.42 0.46 3.09 4.14 .002 .025 Age � 0.42 0.15 � 2.85 0.66 .004 .030 Family structure 0.61 0.64 0.96 1.85 .337 .588 SES 0.08 0.24 0.33 1.08 .743 .774

Individual Internet use 0.20 0.21 0.93 1.22 .353 .588 Effortful control 0.30 0.40 0.75 1.34 .453 .708 Sensation-seeking � 0.10 0.16 � 0.64 0.90 .525 .709 Anger/frustration � 0.21 0.31 � 0.70 0.81 .484 .709 Social skills 0.19 0.15 1.28 1.20 .202 .505 Moral disengagement

0.56 0.39 1.42 1.74 .156 .433

Self-esteem � 0.18 0.40 � 0.44 0.84 .658 .768 Depressive symptoms

1.89 0.67 2.81 6.63 .005 .030

Delinquency 1.03 0.91 1.14 2.81 .256 .540 Academic performance

� 0.16 0.26 � 0.62 0.85 .539 .709

Family Interparental conflict

0.92 0.33 2.74 2.50 .006 .030

Warmth and acceptance

0.70 0.40 1.78 2.02 .075 .289

Behavioral control 0.14 0.33 0.44 1.15 .657 .768 Psychological control

� 0.01 0.33 � 0.02 0.99 .983 .983

Parent-child attachment

� 0.42 0.38 � 1.13 0.65 .259 .540

School School climate � 0.46 0.29 � 1.60 0.63 .110 .344

0.06 0.15 0.41 1.07 .680 .768

(continued on next page)

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

8

predictors for different subgroups from multiple risk domains, which can provide suggestions for designing interventions to protect adoles- cents from bullying and victimization.

4.1. Profiles of bullying and victimization

LPA revealed three distinct profiles: uninvolved, high traditional bully-victims, and high cyber bully-victims. Adolescents in the three profiles differed significantly in their exposure to various forms of bullying and victimization, which suggested that not all adolescents experience the same level or form of bullying and victimization. Thus, there is heterogeneity among forms of bullying and victimization.

Specifically, in the present study, uninvolved adolescents (92%) accounted for a large proportion of participants and reported low levels of bullying and victimization, including traditional direct, traditional indirect, and cyber forms. This group was also identified in previous studies and always accounted for the largest proportion, although the specific proportion varied in different samples (e.g., Kochel et al., 2015; Lovegrove et al., 2012). Additionally, the bully-victims group could be

further differentiated into high traditional bully-victims and high cyber bully-victims. Adolescents in the high traditional bully-victims group (6%) experienced high levels of traditional direct bullying and victimi- zation and medium levels of traditional indirect bullying and victimi- zation. The high cyber bully-victims group (2%) was a special profile that had not been identified in previous studies. Students in this group had high levels of cyber bullying and victimization and medium levels of traditional bullying and victimization. Prior research indicated that cyber bully-victims were often involved in traditional bullying and victimization in real life settings as well as online (Gini et al., 2017; Tokunaga, 2010). Thus, there is an overlap between traditional and cyber forms of bullying and victimization (Juvonen & Gross, 2008). This may be why cyber bully-victims also experience moderate levels of traditional bullying and victimization. The identification of this group supports the possibility that there is heterogeneity among bully-victims and reveals a new important target for prevention and intervention, which is different from high traditional bully-victims and may have special risk or protective factors. However, the size of the high cyber bully-victims group was relatively low. In fact, the prevalence of cyber bullying and victimization have not achieved a consistent conclusion based on previous studies. Similar to our results, in a sample of approximately 440,000 students in the US, Olweus (2013) suggested the prevalence of cyber bullying and victimization near 3%, which was at 25–35% of traditional bullying and victimization. However, Kowalski et al.’s (2014) meta-analysis of cyber bullying suggested that the prev- alence ranged from 10% to 40%. Additionally, they pointed out that the prevalence depended on the background and samples of the study as well as the definition and measurement of cyber bullying. Compared to adolescents in western countries, Chinese adolescents spend more time in school (Xie, Xie, Yang, Bear, & Ling, 2016). Chinese middle and high schools often prohibit students from carrying and using electronic products in schools, thus, students who obey school rules will not have much access to the Internet in schools. In addition, Chinese parents often attach great importance to academic performance and limit their chil- dren’s online time. Based on data from large surveys in China and America, Chinese adolescents spend less time on the Internet than American adolescents (Chinese Academy of Social Sciences, 2018; Twenge, Martin, & Spitzberg, 2019). The less online time may contribute to the lower prevalence of cyber bullying and victimization (Twyman, Saylor, Taylor, & Comeaux, 2010). In fact, Xie et al. (2016) found that the prevalence of cyber bullying and victimization among Chinese adolescents was 4%, which was significantly lower than that of American adolescents. Given that cyberspace may be an extension of school grounds, and electronic devices are just one of the tools used for bullying (Juvonen & Gross, 2008), the prevalence of cyber bullying and victimization should not be overestimated.

Unlike previous studies, we did not identify profiles of only bullies or victims (e.g., Kochel et al., 2015; Lovegrove et al., 2012). One reason for this is that victims may join in on bullying, thereby becoming bully-victims. According to social cognitive theory (Bandura, 1986), observations of bullying behaviors may be associated with potential adoption of such behaviors. Victims who have more experience with bullying may imitate bullying behaviors (Skrzypiec, Askell-Williams, Slee, & Lawson, 2018), especially during adolescence they gradually gain more power to retaliate or join in on bullying. Another possible reason may be that victims seek revenge on bullies. Studies indicated that revenge was positively related to bullying and victimization (Sar- ıçam & Çetinkaya, 2018). Thus, bullying interactions are reciprocal in nature, and former victims may seek revenge on bullies, thus becoming bully-victims. Furthermore, unilateral bullying events without recipro- cation are more likely to gain teachers’ or parents’ attention and be prevented.

It is hard to directly compare our findings with those of prior studies, due to different indicators and samples. However, a few studies have partly supported our findings. Ettekal and Ladd (2017) conducted LPA based on physical, verbal, and relational aggression and victimization,

Table 4 (continued )

Logit SE t OR p p- adjusted

School connectedness Teacher support � 0.09 0.23 � 0.38 0.92 .707 .768 Classmate support � 0.19 0.18 � 1.04 0.83 .299 .575

Peer Peer attachment � 0.60 0.35 � 1.75 0.55 .081 .289 Deviant peer affiliation

1.20 0.35 3.41 3.31 .001 .025

Profile 3 vs. Profile 2 Demographics

Gender � 0.12 0.54 � 0.22 0.89 .828 .914 Age � 0.03 0.16 � 0.20 0.97 .841 .914 Family structure 0.24 0.70 0.34 1.26 .738 .879 SES 0.09 0.26 0.35 1.09 .728 .879

Individual Internet use 0.20 0.22 0.92 1.22 .358 .814 Effortful control 0.55 0.44 1.25 1.73 .210 .583 Sensation-seeking � 0.08 0.19 � 0.40 0.93 .691 .879 Anger/frustration � 0.45 0.33 � 1.36 0.64 .173 .583 Social skills 0.24 0.17 1.43 1.27 .153 .583 Moral disengagement

0.29 0.45 0.65 1.34 .517 .867

Self-esteem � 0.49 0.48 � 1.02 0.61 .306 .765 Depressive symptoms

0.52 0.80 0.64 1.68 .520 .867

Delinquency � 0.43 0.67 � 0.65 0.65 .515 .867 Academic performance

� 0.002 0.28 � 0.01 1.00 .995 .995

Family Interparental conflict

0.89 0.37 2.41 2.45 .016 .200

Warmth and acceptance

0.92 0.44 2.10 2.51 .036 .288

Behavioral control � 0.01 0.37 � 0.02 0.99 .987 .995 Psychological control

� 0.13 0.38 � 0.35 0.88 .729 .879

Parent-child attachment

� 0.68 0.45 � 1.51 0.51 .131 .583

School School climate � 0.66 0.33 � 2.00 0.52 .046 .288 School connectedness

0.06 0.18 0.34 1.06 .735 .879

Teacher support � 0.17 0.24 � 0.71 0.84 .476 .867 Classmate support 0.10 0.20 0.47 1.10 .636 .879

Peer Peer attachment � 0.89 0.37 � 2.40 0.41 .016 .200 Deviant peer affiliation

0.53 0.40 1.30 1.69 .194 .583

Note. Profile 1 ¼ uninvolved group; Profile 2 ¼ high traditional bully-victims group; Profile 3 ¼ high cyber bully-victims group.

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

9

and identified (overall) aggressive-victims and relational aggressive-victims in both middle and high school students. Addition- ally, Giang and Graham (2008) conducted LCA based on the same in- dicators used by Ettekal and Ladd, and identified two separate latent groups for aggressive-victims: highly-victimized aggressive-victims and highly-aggressive aggressive-victims. These findings supported that aggressive-victims is a heterogeneous group; however, these two studies did not include cyber bullying and victimization.

Overall, the present study suggested that adolescents involved in bullying and victimization almost always belonged to the bully-victims group, which was heterogeneous and could be divided into distinct subgroups. It is necessary to examine theoretical subgroups combined by different forms of bullying and victimization in empirical studies. The present study also revealed the overlap between traditional and cyber bullying and victimization, as high levels of cyber bullying and victim- ization were almost always accompanied by medium levels of tradi- tional bullying and victimization. Given that cyber bullying and victimization have unique features and a large proportion of adolescents routinely use the Internet, we should pay attention to cyber forms and the high cyber bully-victims group in future research and intervention design. However, researchers and social workers should also be aware that traditional bullying and victimization are still the most prevalent and serious problems. Thus, we should not overestimate the prevalence and severity of cyber bullying and victimization; otherwise, the focus of prevention and intervention may deviate too much from traditional bullying and victimization (Olweus, 2013).

4.2. Predictors of bullying and victimization profiles

To explore the predictors of distinct profiles of bullying and victim- ization, we selected multiple risk factors from five domains based on Bronfenbrenner’s (1979) human ecology theory. The findings suggested that various individual and social environmental factors significantly predicted profiles of bullying and victimization.

First, among demographic factors, gender and age could differentiate profile members of high traditional and cyber bully-victims from unin- volved adolescents. Our findings are consistent with previous studies that indicate that boys and younger students were more likely to be involved in traditional bullying and victimization (Arseneault et al., 2010; Bradshaw et al., 2013; Murphy et al., 2017). Furthermore, our findings challenged the previous view that cyber bullying and victimi- zation are equally likely to occur regardless of age or gender (Tokunaga, 2010). It is consistent with prior meta-analysis that boys were more involved in cyber bullying and victimization (Barlett & Coyne, 2014). However, unlike Kowalski et al.’s (2014) meta-analysis, we found that younger students were more involved in cyber bullying and victimiza- tion. One possible reason may be the age range. Kowalski et al.’s (2014) review included middle school, high school, and college students, while our study focused only on middle and high school students. Chinese middle and high students are always restricted from using electronic products in schools. Thus, our findings about age difference should not be compared directly with those in other research. Even so, we cannot ignore girls and senior students because they can still potentially be involved in bullying and victimization.

Second, among individual factors, depressive symptoms could pre- dict profile members of high traditional and cyber bully-victims compared to uninvolved adolescents. According to the acting out model, depressive symptoms may be expressed behaviorally through aggression, rule breaking, and other maladaptive behaviors (Wolff & Ollendick, 2006); therefore, adolescents with higher levels of depressive symptoms may be more involved in bullying and victimization. Longi- tudinal studies found that depressive symptoms could predict subse- quent involvement in traditional and cyber bullying and victimization (Bilsky et al., 2013; Yang et al., 2013). Other individual factors did not significantly predict distinct group memberships; however, this does not mean they do not impact bullying and victimization. In fact, in our

initial bivariate analyses (not reported here), most of the individual factors were significantly associated with profile membership. Thus, they only do not make unique contributions to profile memberships. These findings suggest that some individual factors may overlap with other factors and/or their contribution may be mediated by other fac- tors. For example, previous studies have shown that low self-esteem was significantly related to subsequent depression (Sowislo & Orth, 2013); thus, depressive symptoms might mediate the association between self-esteem and bullying and victimization.

Third, among family factors, adolescents who experienced inter- parental conflicts were more likely to be high cyber bully-victims than to be uninvolved or high traditional bully-victims. For these adolescents, they learned conflictual interpersonal communications from their par- ents (Kretschmer et al., 2015), and might use such maladaptive communication methods in social networks, leading to cyber bullying and victimization. Furthermore, cyberspace might be a place where they can escape family conflicts. Compared to high traditional bully-victims, they may spend more time on the Internet engaging in cyber bullying and victimization. We were also surprised to find that warm parenting could predict high cyber bully-victims, when compared to high tradi- tional bully-victims. High-warmth parents have been shown to be more likely to employ permissive or authoritative parenting styles (Maccoby & Martin, 1983). Students from permissive families often have very high levels of Internet use (Valcke, Bonte, Wever, & Rots, 2010), and such frequent Internet exposure might give way to high cyber bullying and victimization. Students with authoritative parents have been shown to have generally better social skills (Zhou et al., 2002), and tend to use cyber bullying because it is harder to detect than traditional bullying.

Fourth, among school factors, negative school climate could predict profile members of the high cyber bully-victims group compared to the high traditional bully-victims group. Previous studies also found that students who perceived a negative school climate were more likely to be involved in cyber bullying and victimization, even after controlling for traditional bullying and victimization (Holfeld & Leadbeater, 2017). Students who perceive a negative school climate may consider school as an insecure place and feel they do not belong there (Wang & Degol, 2016). As such, they may spend more time on the Internet and be more likely to be involved in cyber bullying and victimization. Additionally, we also found that adolescents who had higher levels of classmate support were more likely to be uninvolved students, which was consis- tent with findings from previous studies (e.g., Flaspohler et al., 2009; Yubero et al., 2010). These adolescents do not need to increase their social support and status through bullying behaviors (Pouwels et al., 2017). Further, they were also rarely bullied, presumably because bullies might fear isolation and retaliation from peer supporters.

Lastly, among peer factors, deviant peer affiliation could predict profile members of the high-level traditional and cyber bully-victims groups, when compared to the uninvolved group. Adolescents gener- ally tend to affiliate with peers who are similar to them based on homophily selection (Zhu et al., 2016). Thus, adolescents who make friends with deviant peers very frequently show similar maladaptive behaviors (e.g., aggression) and tend to share similar involvement in bullying with their deviant peers (Espelage, Green, & Wasserman, 2007). These adolescents may bully others to gain friendships from deviant peers, while they may also be bullied by other deviant peers at the same time. This may indicate why adolescents who show high levels of deviant peer affiliation are more likely to be involved in bullying and victimization and become bully-victims. Additionally, we also found that adolescents who had lower levels of peer attachment were more likely to be high cyber bully-victims, compared to high traditional bully-victims. These adolescents are more likely to perceive their peer relationships as poor and less likely to interact with peers (DeMonchy, Pijl, & Zandberg, 2004). Therefore, they may spend more time on social networks and socialize more with online strangers to make up for a lack of friendships. Thus, they are more likely to be involved in cyber bullying and victimization compared to adolescents who have higher

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

10

quality friendships. Additionally, we found some changes after adjustments of p-values,

especially for the prediction of Profile 3 and Profile 2. In fact, statisti- cians have not reached a consistent conclusion on when to perform multiple testing correction. On the one hand, the results of multiple testing without adjustments might have relatively high type I error rate. On the other hand, some statisticians argued that it is not necessary to adjust p-value in exploratory research (e.g., Althouse, 2016; Matsunaga, 2007; Rothman, 1990). For exploratory research undertaken single tests of several different hypotheses, the increased probability is distributed across the entire collection of hypotheses tested rather than localized to any one specific hypothesis (Rubin, 2017). Corrections for multiple comparisons may also reduce statistical power (Rothman, 1990). Given the small sample size of Profile 2 and Profile 3, there was greater risk of making type II error. Since both views were rational, we followed the suggestions of Althouse (2016), (a) describe what was done in the study; (b) report the results and give explanations; and (c) let readers interpret the results according to the characteristics of the current research and the rationality of the results, rather than the researchers to choose which results were correct arbitrarily according to the significance of p-values. Lastly, further studies are needed to confirm the results.

4.3. Strengths, limitations, and future directions

This study has several strengths. First, we collected data from a large sample of 1,529 adolescents, which was relatively diverse and repre- sentative. Second, we included bullying and victimization and tradi- tional and cyber forms simultaneously, which could reveal the status of such phenomenon comprehensively. Third, we included 25 typical risk and protective factors based on theories and empirical research. Furthermore, a person-centered approach was used to reveal the profiles of traditional and cyber bullying and victimization and the unique risk and protective factors for each profile.

Several limitations of our study are also worth noting. First, the present study was based on cross-sectional data. Thus, we cannot draw any causal conclusions from our findings. Prospective longitudinal studies are needed to examine the stability of the profiles over time and the dynamic development of bullying and victimization. Further studies also need to examine the longitudinal associations between risk/pro- tective factors and profiles of bullying and victimization. Second, all measurements in this study were based on adolescent self-reports. Although adolescents are reliable reporters of their information and experiences, self-report measures may be influenced by social desir- ability and common method bias (Moses & Williford, 2017). While we employed various methods (e.g., anonymity, reverse scoring, and clear and accurate expression) to minimize common method bias (Podsakoff, Mackenzie, Lee, & Podsakoff, 2003), a multi-informant approach for data collection, such as peer nomination and teacher-report, is needed in further studies. Third, the profiles identified in this study were based on a specific sample, and therefore might not be generalized to other populations. Further studies should be conducted in other contexts and within different racial groups. Forth, we only focused on microsystems (i.e., individual, family, school, and peer), which are particular com- ponents of the social-ecological model. Further studies should pay more attention to macrosystem factors in the model, such as culture and social environment. Finally, this study used the ecological theory as a frame- work to select risk and protective factors. However, this theory was relatively broad. Further studies should apply specific theories (e.g., frustration-aggression theory, general aggression model) to explore whether the factors in these theories could predict bullying and victimization profiles.

4.4. Implications and conclusion

Our findings have significant implications for the prevention of and intervention for adolescent bullying and victimization. As adolescents

differ in the extent and forms of bullying and victimization exposure, tailored prevention and intervention programs should be conducted (Cook et al., 2010). Specifically, the fact that no sole bullies or victims were identified in our study suggests that bullying behaviors are often accompanied by victimized experience. Thus, we should not only reduce students’ bullying behavior, but also help those who were victimized to get out of the shadow of victimization and learn to protect themselves in reasonable ways, rather than revenge. In addition, as high cyber bullying and victimization are accompanied by traditional bullying and victimization, the involvement of cyber bullying and victimization usually represents relatively serious bullying and victimization experi- ence. Although the proportion of cyber bully-victims is relatively low, they always face higher risks of maladjustment (Cross, Lester, & Barnes, 2015). Therefore, identification and intervention for this subgroup are urgent and necessary. On the basis of traditional interventions, parents and educators can teach adolescents about netiquette and Internet safety to decrease cyber bullying and victimization (Chen et al., 2016; Guo, 2016; Olweus, 2013).

Furthermore, multiple risk and protective factors from ecological subsystems could predict profile memberships. Existing prevention programs are often conducted at the school level such as establishing school rules about bullying, providing students knowledge about bullying and victimization, and monitoring and stopping bullying be- haviors (Hong & Espelage, 2012). Although these anti-bullying practices are important, a multipronged approach that targets ecological com- ponents may be important for addressing bullying and victimization (Cook et al., 2010; Hong & Espelage, 2012). Home-school collaboration is such an approach to prevent and intervene in bullying and victimi- zation (Sheridan, Warnes, & Dowd, 2004). Parents and school workers can communicate and work together utilizing a two-way exchange of information (e.g., daily report cards, school-to-home notes; Cox, 2005). Given that there are a number of factors in ecological systems, selecting the relatively salient individual and contextual factors to design pre- vention and intervention programs may be necessary (Cook et al., 2010). Among individual characteristics, depressive symptoms could predict both high traditional and high cyber bully-victims. Thus, educators and parents should strengthen communications with adolescents and be aware of their psychological state, and psychological services should be provided if necessary. For family environment, reducing interparental conflict and encouraging parents to adopt a warm but supervised parenting style can protect adolescents from bullying and victimization. Skill training, psychoeducation, and cognitive behavioral therapy can be used to improve couple relationships and encourage them to adopt positive parenting styles (Harold & Sellers, 2018). For school environ- ment, creating positive school climate may help reduce bullying and victimization. School personnel can promote communication and emotional connection between teachers and students, and encourage students to help and cooperate with each other (Li et al., 2015). More- over, parents and teachers should provide adolescents with knowledge about peer affiliation and help them build positive friendships. Inter- vention based on ecological subsystems can also help reduce cyber bullying and victimization. For instance, “Cyber Friendly Schools Proj- ect” (Cross et al., 2015) was based on five levels of microsystem (i.e., individual, family, school, peer, and online). Our finding adds to this framework by highlighting the role of depressive symptoms, interpar- ental conflict, and peer affiliation.

In summary, our findings highlight the importance of bullying and victimization profiles and their associations with multiple risk/protec- tive factors. We illustrate the utility of person-centered approaches for characterizing bullying and victimization and the predictive role played by multiple risk and protective factors. Thus, the present study can contribute to a deeper understanding of complex patterns of bullying and victimization as well as help to inform intervention programs designed to reduce adolescents’ bullying and victimization.

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

11

CRediT authorship contribution statement

Yue Ding: Conceptualization, Formal analysis, Writing - original draft, Writing - review & editing. Dongping Li: Conceptualization, Methodology, Supervision, Writing - review & editing, Validation. Xian Li: Writing - review & editing, Validation. Jiale Xiao: Formal analysis, Writing - review & editing, Validation. Haiyan Zhang: Writing - review & editing, Validation. Yanhui Wang: Writing - review & editing, Validation.

References

Althouse, A. D. (2016). Adjust for multiple comparisons? It’s not that simple. The Annals of Thoracic Surgery, 101, 1644–1645.

Antoniadou, N., Kokkinos, C. M., & Fanti, K. A. (2019). Traditional and cyber bullying/ victimization among adolescents: Examining their psychosocial profile through latent profile analysis. International Journal of Bullying Prevention, 1, 85–98.

Arseneault, L., Bowes, L., & Shakoor, S. (2010). Bullying victimization in youths and mental health problems: ‘Much ado about nothing’? Psychological Medicine, 40, 717–729.

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice Hall.

Bandura, A., Barbaranelli, C., Caprara, G. V., & Pastorelli, C. (1996). Mechanisms of moral disengagement in the exercise of moral agency. Journal of Personality and Social Psychology, 71, 364–374.

Bao, Z., Li, D., Zhang, W., & Wang, Y. (2015). School climate and delinquency among Chinese adolescents: Analyses of effortful control as a moderator and deviant peer affiliation as a mediator. Journal of Abnormal Child Psychology, 43, 81–93.

Barker, E. D., Arseneault, L., Brendgen, M., Fontaine, N., & Maughan, B. (2008). Joint development of bullying and victimization in adolescence: Relations to delinquency and self-harm. Journal of the American Academy of Child & Adolescent Psychiatry, 47, 1030–1038.

Barlett, C., & Coyne, S. M. (2014). A meta-analysis of sex differences in cyber-bullying behavior: The moderating role of age. Aggressive Behavior, 40, 474–488.

Bauman, S. (2009). Cyberbullying in a rural intermediate school: An exploratory study. The Journal of Early Adolescence, 30, 803–833.

Bilsky, S. A., Cole, D. A., Dukewich, T. L., Martin, N. C., Sinclair, K. R., Tran, C. V., et al. (2013). Does supportive parenting mitigate the longitudinal effects of peer victimization on depressive thoughts and symptoms in children? Journal of Abnormal Psychology, 122, 406–410.

Bradshaw, C. P., Sawyer, A., & O’Brennan, L. (2009). A social disorganization perspective on bullying-related attitudes and behaviors: The influence of school context. American Journal of Community Psychology, 43, 204–220.

Bradshaw, C. P., Waasdorp, T. E., & Johnson, S. L. (2015). Overlapping verbal, relational, physical, and electronic forms of bullying in adolescence: Influence of school context. Journal of Clinical Child and Adolescent Psychology, 44, 494–508.

Bradshaw, C. P., Waasdorp, T. E., & O’Brennan, L. M. (2013). A latent class approach to examining forms of peer victimization. Journal of Educational Psychology, 105, 839–849.

Brewer, G., & Kerslake, J. (2015). Cyberbullying, self-esteem, empathy and loneliness. Computers in Human Behavior, 48, 255–260.

Bronfenbrenner, U. (1979). The ecology of human development. Cambridge, MA: Harvard University.

Bronfenbrenner, U. (1986). Ecology of the family as a context for human development: Research perspective. Developmental Psychology, 22, 723–742.

Chen, L., Ho, S. S., & Lwin, M. O. (2016). A meta-analysis of factors predicting cyberbullying perpetration and victimization: From the social cognitive and media effects approach. 19 pp. 1194–1213). New Media and Society.

Chinese Academy of Social Sciences. (2018). 2018 Chinese adolescents’ Internet use and network security reports. China, Beijing: Chinese Academy of Social Sciences.

Chi, L., & Xin, Z. (2003). The revision of children’s perception of marital conflict scale. Chinese Mental Health Journal, 17, 554–556.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.

Cook, C. R., Williams, K. R., Guerra, N. G., Kim, T. E., & Sadek, S. (2010). Predictors of bullying and victimization in childhood and adolescence: A meta-analytic investigation. School Psychology Quarterly, 25, 65–83.

Cox, D. D. (2005). Evidence-based interventions using home-school collaboration. School Psychology Quarterly, 20, 473–497.

Cross, D., Barnes, A., Papageorgiou, A., Hadwen, K., Hearn, L., & Lester, L. (2015). A social-ecological framework for understanding and reducing cyberbullying behaviours. Aggression and Violent Behavior, 23, 109–117.

Cross, D., Lester, L., & Barnes, A. (2015). A longitudinal study of the social and emotional predictors and consequences of cyber and traditional bullying victimisation. International Journal of Public Health, 60, 207–217.

DeMonchy, M., Pijl, S., & Zandberg, T. (2004). Discrepancies in judging social inclusion and bullying of pupils with behaviour problems. European Journal of Special Needs Education, 19, 317–330.

Earl, R. M., & Burns, N. R. (2009). Experiences of peer aggression and parental attachment are correlated in adolescents. Personality and Individual Differences, 47, 748–752.

Ellis, L. K., & Rothbart, M. K. (2001). Revision of the early adolescent temperament questionnaire. Minneapolis, Minnesota: Poster presented at the Biennial Meeting of the Society for Research in Child Development.

Erginoz, E., Alikasifoglu, M., Ercan, O., Uysal, O., Alp, Z., Ocak, S., et al. (2013). The role of parental, school, and peer factors in adolescent bullying involvement: Results from the Turkish HBSC 2005/2006 Study. Asia-Pacific Journal of Public Health, 27, 1591–1603.

Espelage, D. L., Green, H. D., & Wasserman, S. (2007). Statistical analysis of friendship patterns and bullying behaviors among youth. New Directions for Child and Adolescent Development, 118, 61–75.

Espelage, D. L., & Swearer, S. M. (2010). A social-ecological model for bullying prevention and intervention: Understanding the impact of adults in the social ecology of youngsters. In S. R. Jimerson, S. M. Swearer, & D. L. Espelage (Eds.), Handbook of bullying in schools: An international perspective (pp. 61–72). NY: Routledge.

Espelage, D. L. (2014). Ecological theory: Preventing youth bullying, aggression, and victimization. Theory Into Practice, 53, 257–264.

Ettekal, I., & Ladd, G. W. (2017). Developmental continuity and change in physical, verbal, and relational aggression and peer victimization from childhood to adolescence. Developmental Psychology, 53, 1709–1721.

Evans, G. W., Li, D., & Whipple, S. S. (2013). Cumulative risk and child development. Psychological Bulletin, 139, 1342–1396.

Flaherty, B. P., & Kiff, C. J. (2012). Latent class and latent profile models. Data analysis and research publication. In H. Cooper, P. M. Camic, D. L. Long, A. T. Panter, D. Rindskopf, & K. J. Sher (Eds.). APA handbook of research methods in psychology (Vol. 3, pp. 391–404). Washington, DC: American Psychological Association.

Flaspohler, P. D., Elfstrom, J. L., Vanderzee, K. L., Sink, H. E., & Birchmeier, Z. (2009). Stand by me: The effects of peer and teacher support in mitigating the impact of bullying on quality of life. Psychology in the Schools, 46, 636–649.

Giang, M. T., & Graham, S. (2008). Using latent class analysis to identify aggressors and victims of peer harassment. Aggressive Behavior, 34, 203–213.

Gini, G., Card, N. A., & Pozzoli, T. (2017). A meta-analysis of the differential relations of traditional and cyber-victimization with internalizing problems. Aggressive Behavior, 44, 185–198.

Gini, G., Pozzoli, T., & Hymel, S. (2014). Moral disengagement among children and youth: A meta-analytic review of links to aggressive behavior. Aggressive Behavior, 40, 56–68.

Glickman, M. E., Rao, S. R., & Schultz, M. R. (2014). False discovery rate control is a recommended alternative to Bonferroni-type adjustments in health studies. Journal of Clinical Epidemiology, 67, 850–857.

G�omez-Ortiz, O., Rey, R. D., Casas, J., & Ortega-Ruiz, R. (2014). Parenting styles and bullying involvement. Culture and Education, 26, 132–158.

Goodman, L. A. (2007). On the assignment of individuals to latent classes. Sociological Methodology, 37, 1–22.

Grych, J. H., Seid, M., & Fincham, F. D. (1992). Assessing marital conflict from the child’s perspective: The children’s perceptions of interparental conflict scale. Child Development, 63, 558–572.

Guo, S. (2016). A meta-analysis of the predictors of cyberbullying perpetration and victimization. Psychology in the Schools, 53, 432–453.

Harold, G. T., & Sellers, R. (2018). Annual research review: Interparental conflict and youth psychopathology: An evidence review and practice focused update. Journal of Child Psychology and Psychiatry, 59, 374–402.

Harter, S. (1988). Manual for the self-perception profile for adolescents. Denver, CO: University of Denver.

Hemphill, S. A., Tollit, M., Kotevski, A., & Heerde, J. A. (2014). Predictors of traditional and cyber-bullying victimization: A longitudinal study of Australian secondary school students. Journal of Interpersonal Violence, 30, 2567–2590.

Hodges, E. V. E., Boivin, M., Vitaro, F., & Bukowski, W. M. (1999). The power of friendship: Protection against an escalating cycle of peer victimization. Developmental Psychology, 35, 94–101.

Holfeld, B., & Leadbeater, B. J. (2017). Concurrent and longitudinal associations between early adolescents’ experiences of school climate and cyber victimization. Computers in Human Behavior, 76, 321–328.

Hong, J. S., & Espelage, D. (2012). A review of research on bullying and peer victimization in school: An ecological system analysis. Aggression and Violent Behavior, 17, 311–322.

Hong, J. S., Kim, D. H., & Piquero, A. R. (2017). Assessing the links between punitive parenting, peer deviance, social isolation and bullying perpetration and victimization in South Korean adolescents. Child Abuse & Neglect, 73, 63–70.

Hong, J. S., Lee, J., Espelage, D. L., Hunter, S. C., Patton, D. U., & Rivers, T. J. (2016). Understanding the correlates of face-to-face and cyberbullying victimization among U.S. Adolescents: A social-ecological analysis. Violence & Victims, 31, 638–663.

Jung, T., & Wickrama, K. A. S. (2008). An introduction to latent class growth analysis and growth mixture modeling. Social and Personality Psychology Compass, 2, 302–317.

Juvonen, J., & Gross, E. F. (2008). Extending the school grounds?—bullying experiences in cyberspace. Journal of School Health, 78, 496–505.

Killer, B., Bussey, K., Hawes, D., & Hunt, C. (2019). A meta-analysis of the relationship between moral disengagement and bullying roles in youth. Aggressive Behavior, 45, 450–462.

Klein, J., Cornell, D., & Konold, T. (2012). Relationships between bullying, school climate, and student risk behaviors. School Psychology Quarterly, 27, 154–169.

Kochel, K. P., Ladd, G. W., Bagwell, C. L., & Yabko, B. A. (2015). Bully/victim profiles’ differential risk for worsening peer acceptance: The role of friendship. Journal of Applied Developmental Psychology, 41, 38–45.

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

12

Kokkinos, C. M., & Antoniadou, N. (2019). Cyber-bullying and cyber-victimization among undergraduate student teachers through the lens of the general aggression model. Computers in Human Behavior, 98, 59–68.

Kokkinos, C. M., Antoniadou, N., Asdre, A., & Voulgaridou, K. (2016). Parenting and Internet behavior predictors of cyber-bullying and cyber-victimization among preadolescents. Deviant Behavior, 37, 439–455.

Kovacs, M. (1992). Children’s depression inventory (CDI) manual. NY: Multi-Health Systems.

Kowalski, R. M., Giumetti, G. W., Schroeder, A. N., & Lattanner, M. R. (2014). Bullying in the digital age: A critical review and meta-analysis of cyberbullying research among youth. Psychological Bulletin, 140, 1073–1137.

Kretschmer, T., Sentse, M., Meeus, W., Verhulst, F. C., Veenstra, R., & Oldehinkel, A. J. (2015). Configurations of adolescents’ peer experiences: Associations with parent- child relationship quality and parental problem behavior. Journal of Research on Adolescence, 26, 474–491.

Lawson, M. A., Alameda-Lawson, T., Downer, J., & Anderson, E. (2013). Analyzing sub- population profiles and risk factors for school bullying. Children and Youth Services Review, 35, 973–983.

Lereya, S. T., Samara, M., & Wolke, D. (2013). Parenting behavior and the risk of becoming a victim and a bully/victim: A meta-analysis study. Child Abuse & Neglect, 37, 1091–1108.

Li, D., He, D., Chen, W., Bao, Z., Wang, Y., & Zhao, L. (2015a). School climate and adolescent problem behaviors: The mediating role of peer victimization. Journal of Psychological Science, 38, 896–904.

Li, Y., Li, D., Li, X., Zhou, Y., Sun, W., Wang, Y., et al. (2018). Cyber victimization and adolescent depression: The mediating role of psychological insecurity and the moderating role of perceived social support. Children and Youth Services Review, 94, 10–19.

Li, D., Li, X., Wang, Y., & Bao, Z. (2016). Parenting and Chinese adolescent suicidal ideation and suicide attempts: The mediating role of hopelessness. Journal of Child and Family Studies, 25, 1397–1407.

Li, D., Li, X., Wang, Y., Zhao, L., Bao, Z., & Wen, F. (2013). School connectedness and problematic Internet use in adolescents: A moderated mediation model of deviant peer affiliation and self-control. Journal of Abnormal Child Psychology, 41, 1231–1242.

Li, D., Li, X., Zhao, L., Zhou, Y., Sun, W., & Wang, Y. (2017). Linking multiple risk exposure profiles with adolescent Internet addiction: Insights from the person- centered approach. Computers in Human Behavior, 75, 236–244.

Little, R. J., & Rubin, D. B. (2002). Statistical analysis with missing data (2nd ed.). NJ: Wiley.

Li, D., Xu, L., Bao, Z., Chen, W., Su, X., & Zhang, W. (2015b). Family financial strain and adolescents’ depression: The effects of perceived discrimination and parent- adolescent attachment. Psychological Development and Education, 31, 342–349.

Li, D., Zhang, W., Li, X., Zhen, S., & Wang, Y. (2010). Stressful life events and problematic Internet use by adolescent females and males: A mediated moderation model. Computers in Human Behavior, 26, 1199–1207.

Lovegrove, P. J., & Cornell, D. G. (2014). Patterns of bullying and victimization associated with other problem behaviors among high school students: A conditional latent class approach. Journal of Crime and Justice, 37, 5–22.

Lovegrove, P. J., Henry, K. L., & Slater, M. D. (2012). Examination of the predictors of latent class typologies of bullying involvement among middle school students. Journal of School Violence, 11, 75–93.

Low, S., & Espelage, D. (2013). Differentiating cyber bullying perpetration from non- physical bullying: Commonalities across race, individual, and family predictors. Psychology of Violence, 3, 39–52.

Maccoby, E. E., & Martin, J. A. (1983). Socialization, personality, and social development: Socialization in the context of the family: Parent-child interaction. In P. H. Mussen, Series, & E. M. Hetherington (Eds.), Handbook of child psychology (4th ed., pp. 1–101). NY: Wiley.

Matsunaga, M. (2007). Familywise error in multiple comparisons: Disentangling a knot through a critique of O’Keefe’s arguments against alpha adjustment. Communication Methods and Measures, 1, 243–265.

Menesini, E., Nocentini, A., & Calussi, P. (2011). The measurement of cyberbullying: Dimensional structure and relative item severity and discrimination. Cyberpsychology, Behavior, and Social Networking, 14, 267–274.

Modecki, K. L., Minchin, J., Harbaugh, A. G., Guerra, N. G., & Runions, K. C. (2014). Bullying prevalence across contexts: A meta-analysis measuring cyber and traditional bullying. Journal of Adolescent Health, 55, 602–111.

Moses, M., & Williford, A. (2017). Individual indicators of self-reported victimization among elementary school-age students: A latent class analysis. Children and Youth Services Review, 83, 33–44.

Murphy, T. P., Laible, D., & Augustine, M. (2017). The influences of parent and peer attachment on bullying. Journal of Child and Family Studies, 26, 1388–1397.

Muth�en, L. K., & Muth�en, B. O. (1998-2015). Mplus user’s guide (7th ed.). Los Angeles, CA: Muth�en & Muth�en.

Nakamoto, J., & Schwartz, D. (2010). Is peer victimization associated with academic achievement? A meta-analytic review. Social Development, 19, 221–242.

Nylund, K. L., Asparouhov, T., & Muth�en, B. O. (2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14, 535–569.

Olweus, D. (1994). Bullying at school: What we know and what we can do? Malden, MA: Blackwell Press.

Olweus, D. (1996). The revised Olweus bully/victim questionnaire. Bergen, Norway: Research center for health promotion. University of Bergen.

Olweus, D. (2013). School bullying: Development and some important challenges. Annual Review of Clinical Psychology, 9, 751–780.

Podsakoff, P. M., Mackenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 879–903.

Pouwels, J. L., Salmivalli, C., Saarento, S., van den Berg, Y. H. M., Lansu, T. A. M., & Cillessen, A. H. N. (2017). Predicting adolescents’ bullying participation from developmental trajectories of social status and behavior. Child Development, 89, 1157–1176.

Raja, S. N., McGee, R., & Stanton, W. R. (1992). Perceived attachments to parents and peers and psychological well-being in adolescence. Journal of Youth and Adolescence, 21, 471–485.

Reijntjes, A., Kamphuis, J. H., Prinzie, P., & Telch, M. J. (2010). Peer victimization and internalizing problems in children: A meta-analysis of longitudinal studies. Child Abuse & Neglect, 34, 244–252.

Rodríguez-Hidalgo, A. J., Pantale�on, Y., & Calmaestra, J. (2019). Psychological predictors of bullying in adolescents from pluricultural schools: A transnational study in Spain and Ecuador. Frontiers in Psychology, 10, 1383–1393.

Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press.

Rothman, K. J. (1990). No adjustments are needed for multiple comparisons. Epidemiology, 1, 43–46.

Rubin, M. (2017). Do p values lose their meaning in exploratory analyses? It depends how you define the familywise error rate. Review of General Psychology, 21, 269–275.

Sarıçam, H., & Çetinkaya, Ç. (2018). Exploring revenge as a mediator between bullying and victimisation in gifted and talented students. Current Issues in Personality Psychology, 6, 102–111.

Schoeler, T., Duncan, L., Cecil, C. M., Ploubidis, G. B., & Pingault, J. B. (2018). Quasi- experimental evidence on short- and long-term consequences of bullying victimization: A meta-analysis. Psychological Bulletin, 144, 1229–1246.

Sheridan, S. M., Warnes, E. D., & Dowd, S. (2004). Home-school collaboration and bullying: An ecological approach to increase social competence in children and youth. In D. L. Espelage, & S. M. Swearer (Eds.), Bullying in American schools: A social- ecological perspective on prevention and intervention (pp. 245–268). Mahwah, NJ: Lawrence Erlbaum Associates Publishers.

Skrzypiec, G., Askell-Williams, H., Slee, P. T., & Lawson, M. J. (2018). Involvement in bullying during high school: A survival analysis approach. Violence & Victims, 33, 563–582.

Solberg, M. E., & Olweus, D. (2003). Prevalence estimation of school bullying with the Olweus bully/victim questionnaire. Aggressive Behavior, 29, 239–268.

Sowislo, J. F., & Orth, U. (2013). Does low self-esteem predict depression and anxiety? A meta-analysis of longitudinal studies. Psychological Bulletin, 139, 213–240.

Stalmach, M., Tabak, I., & Radiukiewicz, K. (2014). Selected family socio-economic factors as predictors of peer violence among school children in Poland. Developmental Period Medicine, 18, 495–505.

Stevens, V., Bourdeaudhuij, I. D., & Oost, P. V. (2002). Relationship of the family environment to children’s involvement in bully/victim problems at school. Journal of Youth and Adolescence, 31, 419–428.

Thapa, A., Cohen, J., Guffrey, S., & Higgins-D’Alessandro, A. (2013). A review of school climate research. Review of Educational Research, 83, 357–385.

Thomas, H. J., Connor, J. P., & Scott, J. G. (2015). Integrating traditional bullying and cyberbullying: Challenges of definition and measurement in adolescents－a review. Educational Psychology Review, 27, 135–152.

Tippett, N., & Wolke, D. (2014). Socioeconomic status and bullying: A meta-analysis. American Journal of Public Health, 104, 48–59.

Tokunaga, R. S. (2010). Following you home from school: A critical review and synthesis of research on cyberbullying victimization. Computers in Human Behavior, 26, 277–287.

Twenge, J. M., Martin, G. N., & Spitzberg, B. H. (2019). Trends in US adolescents’ media use, 1976–2016: The rise of digital media, the decline of TV, and the (near) demise of print. Psychology of Popular Media Culture, 8, 329–345.

Twyman, K., Saylor, C., Taylor, L. A., & Comeaux, C. (2010). Comparing children and adolescents engaged in cyberbullying to matched peers. Cyberpsychology, Behavior, and Social Networking, 13, 195–199.

Valcke, M., Bonte, S., Wever, B. D., & Rots, I. (2010). Internet parenting styles and the impact on Internet use of primary school children. Computers and Education, 55, 454–464.

Wang, M. C., & Bi, X. Y. (2018). Latent variable modeling using Mplus. Chongqing, China: Chongqing University Press.

Wang, M. T., & Degol, J. L. (2016). School climate: A review of the construct, measurement, and impact on student outcomes. Educational Psychology Review, 28, 315–352.

Wang, J., Iannotti, R. J., & Luk, J. W. (2012). Patterns of adolescent bullying behaviors: Physical, verbal, exclusion, rumor, and cyber. Journal of School Psychology, 50, 521–534.

Wang, J., Iannotti, R. J., & Nansel, T. R. (2009). School bullying among adolescents in the United States: Physical, verbal, relational, and cyber. Journal of Adolescent Health, 45, 368–375.

Wang, Q., Pomerantz, E. M., & Chen, H. (2007). The role of parents’ control in early adolescents’ psychological functioning: A longitudinal investigation in the United States and China. Child Development, 78, 1592–1610.

Wolff, J. C., & Ollendick, T. H. (2006). The comorbidity of conduct problems and depression in childhood and adolescence. Clinical Child and Family Psychology Review, 9, 201–220.

Wright, M. F., Kamble, S. V., & Soudi, S. P. (2015). Indian adolescents’ cyber aggression involvement and cultural values: The moderation of peer attachment. School Psychology International, 36, 410–427.

Y. Ding et al.

Computers in Human Behavior 111 (2020) 106439

13

Xie, J., Xie, L., Yang, C., Bear, G., & Ling, Y. (2016). A comparative study of bullying victimization in Chinese and American adolescents. Chinese Journal of Clinical Psychology, 24, 706–709.

Yang, S. J., Stewart, R., & Kim, J. M. (2013). Differences in predictors of traditional and cyber-bullying: A 2-year longitudinal study in Korean school children. European Child & Adolescent Psychiatry, 22, 309–318.

Yubero, S., Ovejero, A., & Larranaga, E. (2010). Perceived social support in school and peer victimization during adolescence. Revista de Psicología Social, 25, 283–293.

Zhou, Q., Eisenberg, N., Losoya, S. H., Fabes, R. A., Reiser, M., Guthrie, I. K., et al. (2002). The relations of parental warmth and positive expressiveness to children’s

empathy-related responding and social functioning: A longitudinal study. Child Development, 73, 893–915.

Zhu, J., Yu, C., Zhang, W., Bao, Z., Jiang, Y., Chen, Y., et al. (2016). Peer victimization, deviant peer affiliation and impulsivity: Predicting adolescent problem behaviors. Child Abuse & Neglect, 58, 39–50.

Zimet, G. D., Dahlem, N. W., Zimet, S. G., & Farley, G. K. (1988). The multidimensional scale of perceived social support. Journal of Personality Assessment, 52, 30–41.

Zych, I., Farrington, D. P., & Ttofi, M. M. (2018). Protective factors against bullying and cyberbullying: A systematic review of meta-analyses. Aggression and Violent Behavior, 45, 4–19.

Y. Ding et al.