The sample is made up of 3180 students of Compulsory Secondary Education (CSE) from 16 schools, 11 public schools and 5 subsidized schools, of Asturias (Spain). Of the assessed students, 28.1% are studying first grade of CSE, 26.5% are in second grade, 24.3% are third-grade students, and 21.1% are in fourth grade. Their ages range between 11 and 19 years (M=13.99, SD=1.38). Concerning gender, 48.5% are boys and 51.5% are girls. Table 1 presents the main descriptive statistics of the sample in the variables analyzed in this study.

Ad hoc questionnaire about sociodemographic data and handling of communication technologies. This questionnaire contains 10 items concerning students’ age, gender, grade, and academic performance, as well as use of electronic devices for communication and the frequency with they are used. Age and grade were assessed by means of two fill-in-the-blank items where the respondents were asked to write their age in years and their school year level. Gender and academic performance were assessed with dichotomic items (male student/female student) and the item “I’ve sometimes repeated a course” (Yes/No). The use of electronic devices for communication was assessed with four items -”I have my own cellphone”, “In my free time, I participate in social networks (Tuenti, Facebook or other)”, “In my free time, I use instant messaging programs (Messenger, WhatsApp or other)”, and “I play on-line with other people”, with a dichotomic response format (Yes/No). Usage frequency of Internet for tasks other than homework was assessed with two items (“In general, how many hours a day do you use Internet for tasks other than homework, from Monday to Friday?” and “In general, how many hours a day do you usually use Internet for tasks other than homework over the weekend?”), with a multiple choice format (None/Less than one hour/Between one and two hours/Between two and three hours/More than three hours).

Cybervictimization Questionnaire (CBV) (Álvarez-García, Dobarro, & Núñez, 2015). This instrument has 26 items, each one describing an instance of aggression suffered through cellphone or internet. The students should mark the frequency with which they were the victim of each one of these situations in the past three months, by means of a 4-point Likert-type scale (1=Never, 2=A few times, 3=Often, and 4=Always). Drawing on the classification proposed by Nocentini et al. (2010), the sentences cover four types of cybervictimization: written-verbal (e.g., “I have received calls insulting me or making fun of me”, “They have made fun of me with offensive or insulting comments on the social networks”, or “I have received insults via SMS or instant messaging programs (e.g., WhatsApp)”); visual (e.g., “I have been forced to do something humiliating that they recorded and later diffused to make fun of me”, “They have uploaded trick photos (modified) of me on Internet to hurt me or make fun of me”, or “They have uploaded real compromising photos or videos of me on the Internet without my permission to harm me or make fun of me”); exclusion (e.g., “They kicked me out or did not accept me in the contact list of some chat, social network — e.g., Tuenti — or instant messaging programs — e.g., Messenger, WhatsApp—, without having done anything wrong, just because it was me”, “Someone has not admitted me or has expelled me from his team in on-line games, without having done anything wrong to justify it” or “They agree to ignore me on social networks”); and impersonation (e.g., “They have impersonated me in Twitter, Tuenti,., creating a false user profile — photo, personal data,.— with which they insulted me or ridiculed me”, “They impersonated me on the Internet, publishing comments in my name, as if it were me”, or “Someone has impersonated another person to ridicule me through Internet or cellphone”). Although the design of the questionnaire was based on this theoretical model of four types of cyberaggression, confirmatory factor analyses carried out with 2,490 students of CSE, aged between 11 and 19 years, from Asturias (Spain), showed that this model had a very similar fit to the one-factor model, which was finally preferred in view of its more parsimonious nature. Test reliability, assessed in terms of internal consistency, was adequate (α=.85).

Cuestionario de Factores de Riesgo para la Cibervictimización (FRC [Cybervictimization Risk Factors Questionnaire]; Dobarro & Álvarez-García, 2014). This self-report aims to identify the degree to which the responder has certain habits, is exposed to certain situations, or makes certain self-appraisals that may constitute a risk or protection factor for cybervictimization, according to available prior evidence. It is made up of 34 items, and responders rate the degree to which each statement is true on a 4-point Likert-type response format (1=Completely false, 2=Rather false, 3=Rather true, 4=Completely true). It was validated with 670 students from CSE, between 11 and 19 years, from Asturias (Spain), and the factor analysis yielded a six-factor model. The Factor Training and Support at School refers to the socio-emotional support received from classmates and teachers, as well as the training received at school about co-existence and the risks of Internet (e.g. “In my school, they have explained the risks of Internet and how to prevent them”, “In class, we usually work on activities concerning education in values (the value of friendship, respect,…)”, or “I have a good friend in class, who listens to me and helps me when I have some problem”). The Factor Offline School Victimization refers to violent acts suffered at school, without the mediation of electronic devices (e.g., “Some classmates reject me in games, walks, or recess activities”, “My classmates mock me and laugh at me”, or “Some students of the school have hit me, either in school or outside the school grounds”). The Factor Risk Behaviors includes usage habits with electronic communication devices that make people more susceptible to cyberaggression (e.g., “I have sometimes met someone whom I only knew from Internet”, “I allow other people to upload my photos or videos on the Internet”, or “I usually publish personal information on my social networks: what I’m going to do, where and with whom, personal photos or videos, family photos or videos;.”). The Factor Parental Control refers to the supervision and establishment of limits on the use of Internet by the family (e.g., “My parents limit the contents I have access to on Internet at home with filters on the computer”, “My parents know my lists of contacts”, or “My parents limit my time on the Internet (either by word or by configuring the computer)”). The Factor Self-esteem refers to the respondent's self-rating (e.g., “I like myself the way I am”, “I can do things at least as well as most of my classmates” or “I am proud of what I do”). Lastly, the Factor Shyness-Social Anxiety includes sentences about inhibition and the feeling of discomfort with regard to others, particularly with people one does not know very well (e.g., “I’m shy and not very talkative, except with my friends”, “I find it difficult to meet new people, make friends, start talking with people whom I do not know”, or “I get uptight if I meet an acquaintance on the street”). Reliability, assessed in terms of internal consistency, was as follows: Training and support at school (α=.75), Offline School Victimization (α=.75), Risk Behaviors (α=.54), Parental Control (α=.80), Self-esteem (α=.73), and Shyness and Social Anxiety (α=.70).

By means of simple random sampling, 16 schools were selected from the total of centers financed with public funds (public and subsidized) in which CSE is taught in Asturias. A second list of 16 alternate centers was also selected. We contacted, first by letter and later by telephone, the Directors of the centers, to request their cooperation. Two Directors refused to participate, and were substituted by the first two alternatives. Each board of directors was informed of the objectives and procedures of the study, its voluntary and anonymous nature, and the confidential treatment of the results. The schools managed the request to the parents for authorization of the students to participate in the investigation, by means of passive consent.

The questionnaires were applied in all the schools in the second or third trimester of the school year 2013-2014. Before completing the questionnaire, the students were also informed of the purpose of the study and of the anonymous and confidential nature of the survey. In general, the students had 20minutes to complete the questionnaires, although this was flexible depending on the age and characteristics of the students. The test was applied by the research team to all the groups in each of the centers, during the regular school schedule.

After the data were entered on a spreadsheet, the possible presence of missing values or outliers on the data matrix was examined. Subjects with 5 or more missing or null items in any of the questionnaires were eliminated from the study. After eliminating these participants, the missing values or outliers still present in the two of the Likert-type tests— the CBV Questionnaire and the FRC Questionnaire —were replaced with the sample item means. The polytomic variables “In general, how many hours a day do you use Internet for tasks other than homework from Monday to Friday?” and “In general, how many hours a day do you use Internet for tasks other than homework on the weekends?” were recoded as dichotomic variables (Threehours or less/More than three hours).

The sample was divided into three subgroups according to their degree of cybervictimization. The No-Cybervictimization Group is made up of students who responded Never to all the CBV questionnaire sentences. The Severe Cybervictimization Group includes students who scored higher than percentile 95 on the CBV (raw score ≥ 41). To determine this cut-off point, the prevalence of students who suffer severe cybervictimization according to recently published review studies was taken into account: between 2 and 7% (Garaigordobil, 2011). The Occasional Cybervictimization Group consists of students who have reported suffering from some of the cyberaggression types assessed, but whose total score on the CBV was below percentile 95. In order to appraise the pertinence of identifying different explanatory models for occasional and severe cybervictims, we verified whether the groups established differed in the predictor variables analyzed. For this purpose, we used Pearson's chi-square test for dichotomous variables and the Kruskal-Wallis H for continuous variables (the assumptions to use parametric statistics were not met).

Next, we examined the degree to which each variable analyzed increases or decreases the risk of being an occasional or severe cybervictim, as well as the possible presence of confounding factors, using multinomial logistic regression analysis. For this purpose, firstly, we calculated the unadjusted Odds Ratio (univariate analysis). Subsequently, we calculated the adjusted Odds Ratio of each variable, statistically controlling for the effect of the rest of the variables by including them in the regression model (multivariate analysis). All the analyses were performed with the statistical program SPSS 19.0 for Windows.

The no-cybervictims, occasional cybervictims, and severe cybervictims were statistically and significantly different in all the variables analyzed, except for academic performance and reported parental control (Table 1).

With regard to occasional cybervictimization, the univariate analyses showed that all the independent variables analyzed, except for training and support at school and having repeated a course, had a statistically significant effect on the probability of being an occasional cybervictim (Table 2). Of the significant variables, only self-esteem was a protective factor, whereas the rest were risk factors. When statistically controlling for possible confounding factors by introducing all the independent variables analyzed in the regression model, gender, shyness and social anxiety, owning a cellphone, playing on-line with other people, and using Internet more than three hours a day from Monday to Friday for tasks other than homework ceased to be statistically significant predictors. In contrast, self-esteem continued to be a protective factor; and age, offline school victimization, parental control, performing risk behaviors on Internet, using social networks and instant messaging software, and using Internet for more than three hours a day during the weekend for tasks other than homework continued to be statistically significant risk factors (Table 2). The factors with the greatest predictive capacity were, in this order, offline school victimization, the use of instant messaging programs, and performing risk behaviors on Internet.

Regarding severe cybervictimization, the univariate analyses showed that all the independent variables analyzed, except for gender and parental control, had a statistically significant effect on the probability of being a severe cybervictim (Table 2). Of them, only self-esteem and training and support at school were protective factors. The rest were risk factors. When statistically controlling for possible confounding factors by including all the independent variables analyzed in the regression model, parental control became statistically significant risk factor, whereas self esteem, shyness and social anxiety, training and support at school, and being a repeater ceased to be statistically significant predictors. Age, offline school victimization, and all the technological variables continued to be statistically significant risk factors (Table 2). Offline school victimization was, with a large difference, the greatest risk factor. The increase by one unit in this variable increases by 14.7 the probability of being a severe cybervictim, after controlling for the effect of the rest of the variables.