Maladaptive social behaviour of students in secondary vocational
education
Koerhuis, M.J.C.
Citation
Koerhuis, M. J. C. (2007, February 6). Maladaptive social behaviour of students in secondary vocational education. Retrieved from https://hdl.handle.net/1887/9751
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/9751
Note: To cite this publication please use the final published version (if applicable).
Chapter 3
Construction of the Questionnaire for Maladaptive Social
Behaviour
Abstract
This study addresses the measurement of maladaptive social behaviour of adolescent students in Dutch secondary vocational education, using a self-report questionnaire. A review on literature and instruments that measure antisocial behaviour was conducted to construct categories of maladaptive social behaviour and statement descriptions typical for these categories. In addition observations were done and students and school-staff of vocational education were interviewed about students’ maladaptive social behaviour in school. Subsequently the internal validity of the questionnaire’s statement descriptions and the respective categories of MSB were investigated in three pilot-studies (Npilot1=101, Npilot2=776, Npilot3=1020). A five- factor structure was identified. Finally, confirmatory factor analyses were used to investigate the robustness of the determined structure. The results confirmed the five- factor structure with sufficient fit indices. Implications and directions for future research are discussed.
Keywords: maladaptive social behaviour, adolescents, questionnaire construction.
Introduction
The Questionnaire for Maladaptive Social Behaviour (QMSB) is a self-report instrument designed to assess maladaptive social behaviour (MSB) of students in secondary vocational education. The questionnaire is based on a self-regulation perspective. Recent theories in self-regulation locate the sources of individual functioning neither in static personality traits, such as hardiness, pessimism or oppression, nor in environmental constraints and affirming factors, but in patterns of self-regulation (Boekaerts, 2005). Self-regulation refers to the strategic use of one’s cognitive, motivational and volitional strategies to reach one’s goals. Recent research has shown that students try to pursue their personal goals alongside goals set by the school (Lemos, 1996; Dowson & McInerney, 2001; Hijzen, Vedder & Boekaerts, 2006).
Many students do not adopt the school goals spontaneously or feel that their personal goals are thwarted at school. These students may feel frustrated and react in several ways. Boekaerts and Corno (2005) have argued that some students may have adequate self-regulative skills at their disposal. They may delay their personal needs and conform themselves to their teachers’ wishes. At such a point, students may seek support from other students or discuss the problem with a teacher. Students who do not have easy access to adequate self-regulation strategies may direct their behaviour toward their personal goals only, for example chatting for fun, or making fun of a teacher to increase their popularity. Other students may try to control their emotions by vandalizing things in and around school, or calling names at a teacher or other students. These examples show that MSB can take several forms, but in essence can be regarded as an outcome of an interaction process between the students’ personal values, needs and interests (higher order goals) and characteristics and features of the environments. When more goal-frustration is experienced by the students, it will become increasingly harder to regulate their behaviour in an adequate way.
In Dutch secondary vocational education, several incidents of students’
misbehaviour have led to all kinds of safety precautions in and around school, for example discussions about implementing detection gates, hiring security employees and implementing identification passes. Several forms of misbehaviour at school occur, for example disturbing the lesson, skipping school and calling names (Neuvel, 2004). This is a serious problem, often co-ocurring by a declined motivation for school.
Need for a new Instrument
In the social sciences, a lot of research is available on antisocial behaviour.
Koerhuis and Boekaerts (in preparation) reviewed the available empirical research on students’ maladaptive behaviour and listed the different operationalizations. The review showed that the majority of the research focused on explaining and predicting delinquent behaviour (e.g., Garnefski, 2000; Aalsma & Lapsley, 2001; Loeber, 1982;
Farrington, 1995; Björkqvist, Lagerspetz, & Kaukiainen, 1992), substance abuse (e.g., Newcomb, Maddahian & Bentler, 1985; Wills, Rensko, Ainette & Mendoza, 2004;
Bingman & Shope, 2004) and violent behaviour (e.g., DiNapoli, 2003; Laufern, Harel, 2003; Williams, Mulhall, Reis & DeVille, 2002). Less attention is given to internalizing
behaviour (e.g., Rubin & Mills, 1988; Asendorf, 1990; Younger & Daniels, 1992), impolite and annoying behaviour (Kowalsky, 2001) and various types of misbehaviour at school. When misbehaviour at school is measured the focus is often on bullying (e.g., O’Moore & Kirkham, 2001; Haynie, Nasel, Eitel, Crump, Saylor, Yu, et. al., 2001), trespassing school rules reflected in truancy and tardiness (e.g., Vazsonyi, &
Pickering, 2003; Talbott, & Thiede, 1999) or on outcomes of misbehaviour, such as being suspended or being called to a principal (e.g., Mortimer, Finch, Shanahan, &
Ryo, 1992; Cochran, & Bo, 1989)
.
Note that the latter operationalization does not even refer to actions undertaken by the adolescents themselves. In line with this focus, the majority of the instruments that are used in a school context assess delinquent behaviour of adolescents (e.g., Self-Reported Delinquency scale, Elliott & Ageton, 1980; Deviant Behaviour Scale, Jessor & Jessor, 1977; Self reported Delinquency Scale, Short & Nye, 1958).In the present study we defined MSB as behaviour that, according to a majority of the reference group (students and staff) in a certain context (school), is undesired or inappropriate. Hence, we do not only consider serious antisocial acts like delinquent behaviour as MSB, but also milder forms like impolite or unfriendly behaviour. It was necessary to develop a new questionnaire, because frequently, existing questionnaires did not assess this wide range of MSB. Furthermore, in the past some researchers assumed that personality traits underlie various types of MSB and instruments reflect this perspective (e.g., Youth Self Report, Achenbach, 1991; Rutter School Behaviour Scale, Rutter, Tizard & Whitmore, 1970). Their aim is to identify adolescents who are at risk, by assessing the deviation from a clinical norm;
respondents are asked how much a statement applies to them, instead of the frequency of occurrence.
Defining MSB as the outcome of an interaction process between the students’
current goals and the context in which they function, it is essential to present the students with descriptions of concrete behaviour in specific school settings. Several such settings can be distinguished, that vary in their physical locations (the classroom, library or outside school), work conditions (schoolwork or homework) and types of interaction between students and staff (during cooperation or social talk). Existing questionnaires to measure adolescents’ MSB often use general statements, for example ‘behaves too young for his/her age’, ‘braggs’ and ‘achieves insufficiently’
(CBCL; Achenbach, 1991).
Another reason why a new instrument is needed is that a great number of the existing questionnaires require parents or teachers to report on students’ maladaptive behaviour (i.e., some frequently used questionnaires in the Netherlands are the Dutch version of the Conners Teacher Rating Scale, Blöte & Curfs, 1986; SCHOBL-R, Bleichrodt, Resing & Zaal, 1993). We prefer to use a self-report instrument, because we expect that most reliable information about this topic should come from the students themselves. Other research has shown that self-report questionnaires give reliable information on antisocial behaviour of adolescents (e.g., Farrington, Loeber, Stouthamer-Loeber & Schmidt, 1996). To measure the frequency of the actions that vocational students engage in at school, respondents need to have excellent knowledge about the behaviour students engage in, both inside and outside the classroom. In most schools for secondary vocational education students remain quite anonymous, because vocational schools are usually large and students see several
teachers a day. Therefore, it is not probable that teachers can give an accurate report of all the students’ MSB at school.
The final reason why many existing questionnaires do not fit our purpose is that questionnaires containing items on misbehaviour at school are often meant for elementary school children (i.e., Bristol Social Adjustment Guide, Stott & Sykes, 1974;
Dutch version of the Conners Teacher Rating Scale, Blöte & Curfs, 1986; SCHOBL-R, Bleichrodt, Resing & Zaal, 1993; CBCL, Achenbach, 1991). We reasoned that, to measure adolescent MSB at school, the statements that are provided need to be suitable and recognizable for the age group under study. Many of the items in the mentioned questionnaires are not suitable for adolescents, for example ‘this student colours outside the lines’ (SCHOBL-R).
In summary, existing instruments to measure MSB are either a) not asking for various types of MSB, b) not asking for concrete examples of school misbehaviour, c) not questioning the frequency of MSB, d) not self-reports questionnaires or e) not suitable for adolescents.
Construction of the Questionnaire for Maladaptive Social Behaviour
In order to address the outlined shortcomings of existing instruments, a questionnaire was developed to measure MSB in educational settings, and it was specifically tuned to the context of secondary vocational education. Henceforth this Questionnaire for Maladaptive Social Behaviour will be referred to as QMSB. To detect salient examples of MSB of students at school, we followed three avenues. First we inspected several reliable scales that were developed to measure substance use, delinquency, and withdrawn behaviour. From these questionnaires ideas were drawn for item construction and types for MSB (i.e., YSR, Achenbach, 1991; Bergen Questionnaire on Antisocial Behaviour, Olweus, 1999; Bristol Social Adjustment Guide, Stott & Sykes, 1974; Conners Teacher Rating Scale, Blöte & Curfs, 1986;
SCHOBL-R, Bleichrodt, Resing & Zaal, 1993). Second, observations were done in classrooms, a library, a restaurant and other areas at school. Observations showed for example that students were disturbing others by chatting in class and in the library, making a mess in and around school, and provoking other students. Third, we interviewed students and staff in vocational schools. Individual interviews were held with three students and 9 staff members (6 teachers, 2 security agents, 1 library employee). The individual interviews revealed ‘function related’ examples of MSB.
Security staff for example mentioned mobile phones or money being stolen and fights between students, library staff mentioned tearing pages out of books. Teachers were most bothered by disruptive behaviour in class, including disrespectful behaviour and maladaptive behaviour concerning schoolwork (e.g., not making homework, handing work in too late). The students themselves mentioned examples of provoking behaviour of other students and passive behaviour during cooperation tasks. Group interviews were held to discuss with the students whether or not they considered the formulated examples of maladaptive behaviour indeed as “maladaptive”. These interviews were held with alternating groups of students with an average number of four. In total, twenty students participated. When consensus was reached on the deviancy of a behavioural statement, it was taken further in the process. Finally, 77 items were selected based on existing questionnaires, observations and interviews.
The items covered a wide range of school behaviours. Based on discussion with experts we expected these items to refer to 7 types of behaviour: (1) Delinquent Behaviour, (2) Impolite Behaviour, (3) Maladaptive Classroom Behaviour, (4) Maladaptive Schoolwork Behaviour, (5) Withdrawn Behaviour, (6) Unfriendly Behaviour toward Teachers, and (7) Unfriendly Behaviour toward Students. Eight statements on prosocial behaviour, taken from the social desirability scale of the SCHOBL-R (Bleidchrodt, et al., 1993), were added to ensure students’ concentration during completion of the questionnaire and to detect unreliable answers.
Two studies were conducted to develop a reliable instrument. In the first study three pilot studies were used to test and revise the items and determine the structure of the questionnaire. In the second study the robustness of the structure of the questionnaire was tested in two confirmatory factor analyses.
Both studies were part of a longitudinal research project on the motivational self-regulation of students in secondary vocational education. Participating students were asked, once in their first year and twice during their second year, to fill in a battery of questionnaires, concerning their goal preferences, perception of the school climate and social support, cooperative learning and MSB. The first data-wave was used for the development of the QMSB (during this period the three pilot studies were conducted). The second and third data-wave were used to test the robustness of the structure using confirmatory factor analyses.
Study 1: Exploring the Structure of the QMSB
The study addressed the internal validity of the 77 statement descriptions in seven types on social maladaptive of adolescent students in secondary vocational education. In the questionnaire students were asked to rate the statements on a Likert scale ranging from 1 through 4, thus indicating how often they showed the behaviour (i.e., never, not often, pretty often, very often) during the last three months at school.
Method Population and samples
About 470.000 students attend a course in Dutch secondary vocational education. In this type of education students are trained for executive professions, like car mechanic, health care worker, nurse or baker. Students can follow a course at different levels, ranging from 1 (low) to 4 (high). Courses at level 1 take 1 year, and students from this level usually continue with a course in the second year at level 2.
Because of the longitudinal character of the present study we chose to combine the
data from students in level 1 and 2 and thus create samples from course levels ‘high’
(level 4), ‘medium’ (level 3) and ‘low’ (level 1 and 2) (See Table 1).
To try out the first version of the questionnaire three classes of in total 101 first- year students of the technology department (level four), of one school were selected (Pilot 1). The internal validity of the final version of the QMSB was examined in the second and third pilot study. In the second pilot study, forty-nine classes of in total 776 first-year students in 10 departments were randomly selected from six schools. The mean age of the students was 17 years and 10 months. Their age ranged from 16 to 22 years. Of this sample 44% was male, 48% was females and of 8% the gender was not known. Twenty percent followed a course at a low level, 12% at a medium level, 60% at a high level, and of 8% of the participants this was not known.
In the third pilot study, sixty-nine classes of in total 1020 students were selected from six vocational schools. The mean age of the students was 17 years and 8 months. Most students (about 90 %) were between 16 and 19 years old. In this sample 28% was male, 61% was female and 11% did not report their gender. Fourteen percent followed a course at a low level, 38% at a medium level, 38% at a high level, and of 10% this was not know.
Procedure
During regular lessons the participating students were informed about the objectives of the study. The confidentiality of the results was emphasized. The students voluntarily completed the questionnaire in about 15 minutes and were afterwards invited to comment on the wording of the statements. They were also asked whether they thought the statements covered students’ MSB in school.
Results Pilot Study 1
We computed Cronbach’s alpha coefficients and the corrected item total correlations for each of the seven types of MSB of the 101 participating students in Pilot study 1. The alpha coefficients demonstrated that the statements in each type were sufficiently interrelated (.73 ≤ α ≤ .88). The type Maladaptive Schoolwork Behaviour contained one statement with an item total correlation lower than .20 (sucking up to a teacher to get something done; ITC = .02). An item total correlation higher than .20 is considered sufficient (Streiner & Norman, 1995). This statement was changed into 'acting particularly kind to a teacher to get one’s way’. Furthermore, students suggested rewording of the statement 'using alcohol or soft drugs’ into two separate statements, because many students reported using alcohol, but not drugs.
Pilot Study 2
Using the data collected from the 776 first year students we conducted a principle component analysis followed by varimax rotation. It yielded six factors with an Eigenvalue larger than 1, which explained 50.9 % of the variance of 46 out of 78 statements. Statements with low factor loadings (< .40) or loadings higher than .40 on multiple factors were excluded. Out of the seven predefined seven types of MSB, five factors corresponded: Delinquent Behaviour (11 statements), Withdrawn Behaviour (7 statements), Maladaptive Classroom Behaviour (7 statements), Maladaptive Schoolwork Behaviour (6 statements), and Impolite Behaviour (3 statements).
Contrary to what was expected, the statements in the types Unfriendly Behaviour toward Teachers and Unfriendly Behaviour toward Students loaded on a single factor, comprising of 10 statements. Apparently, both kinds of behaviours refer to a common characteristic, namely behaving unfriendly. As can be seen in Table 1 items in the first factor obviously refer to serious antisocial behaviours. The items in this type refer to using physical violence toward teachers and other students, theft and vandalism.
Withdrawn Behaviour consists of items referring to isolating one’s self from others in an active way (i.e., ‘isolating yourself from others’, ‘sitting apart from others’) and items referring to passive withdrawn behaviour (i.e., ‘not joining discussions in class’, ‘not taking initiatives’). The fourth type ‘Maladaptive Classroom Behaviour’ refers to disruptive behaviour in the classroom (i.e., ‘not participating in the lesson’, ‘distracting other students’). Three out of the seven items in this type also have quite high loadings (> .30) on ‘Unfriendly Behaviour’ or ‘Maladaptive Schoolwork Behaviour’. In the latter type, which consists of items that refer to trespassing school-regulations, items have component loadings between .45 and .59, which is low compared to the other types.
The final type ‘Impolite Behaviour’ is a rest type with three items. Two of these items also have loadings higher than .30 on Maladaptive Classroom Behaviour and Maladaptive Schoolwork Behaviour. We concluded that this type of MSB did not yet meet our criteria.
Careful inspection of the statements that were included and excluded, suggested that slight rewording of twelve excluded statements would be an option. For example, in the type ‘Delinquent Behaviour’ four statements missed a clear reference to the school context (e.g., using drugs, drinking alcohol); two statements referred to maladaptive acts toward students and teachers, thereby excluding other school staff members. In the type ‘Unfriendly Behaviour’ it was expected that stronger wording of two statements would be more compatible with the other statements in this type:
‘Gossip about a student’ and ‘gossip about a teacher’ probably had better changed into ‘blacken a student’s reputation’ and ‘blacken a teacher’s reputation’. Only three out of twelve statements in the type Impolite Behaviour were included in the factor structure. Based on the interviews with students and staff, we expected that several of the excluded statements measured Impolite Behaviour, for example ‘making trash’.
Possibly, the general wording of the statements may not have been accurate for the students in this study and could better be changed into 'leaving your litter, instead of throwing it in a waste paper basket'. Finally, three statements in the type Delinquent Behaviour, seemed to refer to rather similar behaviours (‘(trying) to steal a wallet or money’, ‘taking someone’s bike, moped or car without permission’, and ‘taking
someone’s belongings without permission’). The last statement was retained, because its content comprised the content of both other statements.
We computed Cronbach’s alpha coefficients and the corrected item total correlations in each of the six types of MSB for the total sample and for sub-samples based on gender and course-level. The alpha coefficients demonstrate that the statements in each type are sufficiently related to each other in the total sample (.71 ≤ α ≤ .88), in the sub-sample based on level (.60 ≤ α ≤ .93), and in the sub-sample of males (.73 ≤ α ≤ .89). Relatively low coefficients were found in the type Impolite Behaviour, which could be explained by the low number of statements (3). The lowest alpha coefficient was found in the type Delinquent Behaviour in the sample of females (.55), due to relatively low item total correlations of the statements on stealing.
Nevertheless, these statements were kept in the questionnaire, because interviews had shown that stealing was considered an important example of delinquent behaviour by these students. Furthermore these statements did contribute to the internal validity of the type in the total sample, the sub sample of males, and the samples based on course-level.
The scale for Impolite Behaviour appeared to be weak. Reliability coefficients for the sample and sub-samples varied between .60 and .73. Careful inspection of the statements in the questionnaire that were included and excluded suggests that for better results measurements of MSB of students at these levels of vocational education should refer to rather concrete descriptions of behaviours. In view of the students’ need for rather concrete descriptions, it was expected that the best way to strengthen the measurement of the type Impolite Behaviour was to add concrete statement descriptions, such as: ‘not thanking a student when he/she does something for you’, ‘not thanking a teacher when he/she does something for you’, ‘not apologizing when you are late’, ‘not greeting someone who greets you’, ‘chatting while other people are telling something’, ‘sticking chewing gum at things in school’, ‘blot things in school’ en ‘burping loud on purpose’. In a similar vein, the type Schoolwork Behaviour was to be improved by adding statements such as ‘taking advantage of someone else’s work’, ‘making up excuses about why you didn’t make your schoolwork’ and
‘making up excuses in order to skip a lesson’.
Pilot Study 3
In the final pilot study, using the data of 1026 first year students, principal component analysis followed by varimax rotation yielded five factors with an Eigenvalue larger than 1.
Table 1: Factor Loadings of the Statements in the QMSB. (Loadings suppressed below .30)
Items Component lading
1 2 3 4 5
MSB toward Schoolwork and Rules
Making up excuses why you didn’t finish your work .760
Making up excuses to be able not to go the lesson .699
Handing in work you did not make yourself .678
Playing truant .639
Handing in work too late .636
Losing schoolwork .628
Behaving extra nice to get your way .575
Showing that you are bored .546 .382
Pretending you do not see things .532 Distracting other students .510 .330
Quit working when the teacher leaves .456 .364
Talking when another student gives a presentation .452 Making up stories .445 Not participating in the lesson .438 .328
Jumping the queue .365 Delinquent Behaviour Selling drugs .840
Using harddrugs .833
Beating up a student .795
Hitting a teacher .772
Taking someone’s belongings .650
Damaging belongings of another student .602 .334 Fighting with a group .590
Carrying weapons at school .569
Damaging the school or things outside school .444 .391 Drinking so much at school you get a bit drunk .440
Unfriendly Behaviour Making fun of a student .706
Provoking a student .669
Making fun of a teacher .659
Calling names at a student .649
Provoking a teacher .625
Calling names at a teacher .606
Excluding a student .600
Making sexual remarks .586
Blacken a teacher’s reputation .567
Withdrawn Behaviour Not joining conversations .754
Not taking initiatives in contact with other students .747
Avoiding conversations with other students .730
Not joining discussions in the class .717
Not taking inititiative in cooperation tasks .672
Isolating yourself from other students .611
Avoiding cooperation tasks .604
Sitting apart from other students .545
Impolite Behaviour Not thanking a teacher when he/she does something for you .823
Not thanking a student when he/she does something for you .822
Not apologizing when you are late for class .657
Not greeting when someone greets you .643
These five factors explained 52.8% of the variance of 46 statements (see Table 1): 1) Maladaptive Behaviour toward Schoolwork and Rules (15 statements), 2) Delinquent Behaviour (10 statements), 3) Unfriendly Behaviour (9 statements), 4) Withdrawn Behaviour (8 statements), and 5) Impolite Behaviour (4 statements). In contrast to our expectations, the two types Classroom Maladaptive Behaviour and Schoolwork Maladaptive Behaviour loaded on one factor indicating that these two factors referred to the same construct (i.e., behaviour concerning specific school situations). Moreover, this construct appeared to be the strongest factor in the new version of the QMSB. In Delinquent Behaviour the same items were found as in the previous version. Furthermore, the results show that the statements on alcohol and drugs could be included in Delinquent Behaviour. The types Unfriendly Behaviour and Withdrawn Behaviour comprised the same statements as was expected, with all high loadings on one of the types. Compared to the results that were found in the previous versions of the questionnaire, the results showed that the four revised statements in the type Impolite Behaviour clearly illustrated to the students what is meant by this type of maladaptive social behaviour. We computed Cronbach’s alpha coefficients and the corrected item-total correlations in each of the five types of maladaptive social behaviour for the total sample and for sub-samples based on gender and course-level.
The overall indices of internal consistency demonstrated that the statements in each type were sufficiently related to each other in the total sample (.80 ≤ α ≤ 89), in the sub-sample of males (.79 ≤ α ≤ .91) and females (.73 ≤ α ≤ .86) and in the sub- samples based on course-level (.77 ≤ α ≤ .92). In a last attempt to increase the reliability coefficient of Delinquent Behaviour for girls, the statement ‘threatening someone’ was changed into the more detailed and concrete formulation of ‘threatening someone to hurt him/her’. Furthermore two statements were added to the types Impolite Behaviour (i.e., ‘not thanking someone when he/she gives you something’), and Withdrawn Behaviour (i.e., 'avoiding contacts with (certain) students’).
A final adjustment was made to the anchor points of the 4-point scale. It was changed into a 5-point scale (i.e. never, not often, sometimes, pretty often, and very often). By adding ‘sometimes’ as a middle category we hoped to increase the sensitivity of the instrument. During assessments of the questionnaire several students had indicated that they missed this middle category. An extra desirable effect would be that the variance increases, because the mean-scores and standard deviations of the statements in the five types of social maladaptive behaviour are rather low. Mean scores in this pilot varied from 1.05 for Delinquent Behaviour to 1.79 for MSB toward Schoolwork and Rules. Standard deviations varied form .20 for Delinquent Behaviour to .46 for MSB toward Schoolwork and Rules.
Conclusion
In three pilot studies we explored the structure of the 77 items of QMSB as were formulated based on the literature review, interviews with students and staff and observations in senior vocational education. By rephrasing, adding and deleting several items to improve the structure and validity of the questionnaire, 49 items
referring to five types of MSB, seemed to form a solid structure in the final version.
Internal consistency of the scales is good for the whole sample and for sub samples based on gender and course level. In study 2 we assessed the robustness of the structure of the QMSB.
Study 2: Confirming the Structure of the QMSB
We used Confirmatory Factor Analysis to identify how well the final structure of the questionnaire represents the data gathered in the second and third data wave. The final version of the questionnaire was administered to the students who took part in the study at the second and third data wave of the longitudinal study on self-regulation in vocational schools. Hence, the data sets that were used in study 2 were gathered halfway (T2) and at the end (T3) of the students’ second year. The two data sets were used to investigate the robustness if the structure of the QMSB. More specifically, we examined whether the five types of the QMSB (1) Maladaptive Behaviour toward Schoolwork and Rules, (2) Delinquent Behaviour, (3) Unfriendly Behaviour, (4) Withdrawn Behaviour, and (5) Impolite Behaviour, could be found in the two data sets both for the whole sample and for sub-samples based on gender. Gender is an important variable to take into account, since previous studies have shown differences in prevalence of different types of MSB (i.e., Moffitt, Caspi, Rutter, & Silva, 2001;
Björkqvist, Lagerspetz & Kaukiainen, 1992).
Method Sample
At T2 (January – March) 931 students from 77 classes were selected from 10 schools. Mean age of the students at T2 was 19 years and 2 months. Thirty-one percent was male, 54% was female, and 16% did not report their gender. Seven percent followed a course at a low level, 23% followed a course at a medium level, and 57% followed a course at a high level. Of 13% the course level was not known. At T3 (May-July) 642 student from 46 classes participated, from nine schools. Their mean age was 19 years and 4 months. Nineteen percent was male, 58% was female, and of 3% the gender was missing. Ten percent followed a course at a low level, 22% at a medium level, 40% at a high level and 28% the course level was not known.
Results
Descriptive Statistics
Inspection of item total correlations and Cronbach’s alphas showed that the items were sufficiently related to each other in every type of behaviour at T2.
Cronbach’s alphas ranged from .88 to .97. The lowest Item total correlation was found for Unfriendly Behaviour in students who followed a course at the medium level (.43).
This is still considered sufficient. At T3 Cronbach’s alphas for each type of behaviour ranged from .85 to .96 and the item total correlations showed the items to be sufficiently coherent (> .20) (Streiner & Norman, 1995), except in the type Delinquency for students in the high course group (.12 - .93). Three items (‘carrying weapons at school’, ‘fighting with a group against another group at school’, and ‘threatening someone to hurt him/her’) had an item total correlation lower than .20. This is probably due to the low variance on these items (standard deviations are respectively .31, .12, 17), although Table 2 shows that the standard deviations in these data-waves increased compared to the first data-wave, when a four-point Likert-scale was used.
Confirmatory Factor Analyses
A confirmatory factor analysis was conducted using LISREL 8.54 (Jöreskog, 2004). We wanted to test whether the structure of the QMSB that was found in the pilot data collections at Time 1 represents a good fit for the data at Time 2 and 3.
Analyses were performed on an asymptotic covariance matrix, because raw data did not suffice the assumption of multivariate normality (Browne, 1984). Maximum Likelihood was chosen as a method of estimation. This quite robust method was preferred above Weighted Least Squares (WLS), because WLS requires a very large sample. Maximum Likelihood is considered the best alternative (Jöreskog, 2004).
Satorra-Bentler Chi square statistics were computed, instead of minimum fit function chi-square, because this statistic does not require normality. When necessary the ridge option was used (Wothke, 1993).
A Confirmatory Factor Analysis was conducted on two models. The first model (Figure 1, left side) represents the structure as found in the final version in Study 1: Each of the 49 items loads on one of the five latent variables (1) Maladaptive Behaviour toward Schoolwork and Rules, 2) Delinquent Behaviour, 3) Unfriendly Behaviour, 4) Withdrawn Behaviour and 5) Impolite Behaviour. In the second model one latent variable was added, namely a higher-order overall factor of ‘Maladaptive Social Behaviour’. This model is an alternative to the ‘single-model approach’ as proposed by Jessor and Jessor (1977), which suggests a common factor of MSB. The existence of a common underlying factor has been a topic of interest for many researchers. It is argued that various types of antisocial behaviour (i.e. delinquency, drug use and aggression) all refer to one underlying construct, namely problem behaviour. Other researchers have argued that such an underlying factor can only partially explain the variance of several types of maladaptive behaviour.
Model 1: Model 2:
Figure 1. Models
Although, strong correlations have been found between several types of MSB we reasoned that replacing the separate factors by one common ‘problem behaviour construct’ would throw away too much information. Therefore, we assessed an alternative model that was also found by Farrell, King and White (2000). They assessed the structure of a questionnaire that measured drug use, delinquency and aggression and discovered two models with sufficient and almost equal fit indices; a model with three intercorrelated factors and a model with these three factors (now uncorrelated) and one higher-order factor for problem behaviour. Table 2 shows that the intercorrelations of the types of maladaptive behaviour of the QMSB are quite high (.35 - .65, see Table 4), suggesting a shared construct. However, study 1 showed that five types of MSB can reliably be distinguished. Therefore, a structure with one common factor of MSB, representing all items, does not seem appropriate. Adding a higher-order factor though, could provide a practical construct for the shared variance of the five types of MSB. Hence, in this study we propose a competing model with a higher-order factor for MSB (see Figure 1, Model 2, right sight).
After listwise deletion a sample of 830 students at T2 and 550 at T3 remained and were used in further analyses1. Table 3 shows the chi-square statistics, degrees of freedom, and fit indices. Chi-square and the ratio of chi-square to degrees of freedom are sensitive to sample size. The normed fit index (NFI) and the comparative fit index (CFI) are two fit indices that reflect the improvement of fit of the model, compared to the absence of any model on the data. Values close to .95 are indicators of a good fit (Hu & Bentler, 1999). The root mean square error of approximation (RMSEA) is an indicator of the discrepancy per degree of freedom. A value lower than .05 is an indicator of a close fit, a value up to .08 indicates a reasonable error of approximation and a model with a RMSEA value higher than .08 is considered as insufficient (Browne & Cudek, 1992). All parameters in the model are freely estimated, except for the variance of the latent variables, these were standardized (set to 1.00).
2
3
4 1
5
1
It em s
2
3
4
5
It em s
MSB overall
Table 2: Means, Standard deviations and correlations between types of behaviour (T2 left of the diagonal, T3 right of the diagonal).
Table 3: Fit indices of Model 1 and 2 at T2 and T3 and male and female.
Model Sample χ2 st-bentler p-value Df NFI CFI RMSEA Model 1 T2 ALL (N = 830) 5097.58 <.0001 1117 0.96 0.97 0.066
T2 Male (N = 255) 2680.96 <.0001 1117 0.94 0.95 0.074 T2 Female (N =
446)
3318.35 <.0001 1117 0.97 1.00 0.067 T3 ALL (N = 550) 3632.35 <.0001 1117 0.96 1.00 0.064 Model 2 T2 ALL (N = 830) 5139.99 <.0001 1122 0.96 0.97 0.066 T2 Male (N = 255) 2706.30 <.0001 1122 0.93 0.95 0.075 T2 Female (N = 446) 3375.77 <.0001 1122 0.97 1.00 0.067 T3 ALL (N = 550) 3687.04 <.0001 1122 0.95 1.00 0.063
Inspection of Table 3 informs us that the indices of fit for both models indicate a good fit for the data. The RMSEA (.066) indicate a moderate fit at T2. The difference in fit between the first and the second model was statistically significant (Δχ² = 42.41, Δdf
= 5, p < .01), implying that the second model is a less good representation of the data compared to the first model, despite the fact that the fit indices are exactly equal (on two decimals). Though Chi-square differs significantly between the two models, a substantial difference is not confirmed by the fit indices.
Both models were also tested separately for boys and girls at T2. At T3, the number of boys was too small to conduct a reliable analysis. Results are presented in Table 3. Fit indices indicate a reasonable fit for both groups, and both models seem to fit better for girls.
Conclusion and Discussion
M SD Delinq
uent Behavi
our
Impolit e Behavi
our
Unfrien dly Behavi
our
MSB toward Schoolw ork and Rules
Withdraw n Behaviou
r
T2 T3 T2 T3
Delinquent Behaviour
1.29 1.14 .66 .46 --- .44 .53 .35 .35
Impolite Behaviour
1.65 1.42 .81 .62 .49 ---- .47 .39 .45
Unfriendly Behaviour
1.62 1.37 .74 .51 .63 .63 ---- .59 .39
MSB toward Schoolwork and Rules
2.22 2.06 .66 .60 .55 .49 .63 ---- .38
Withdrawn Behaviour
1.97 1.79 .71 .61 .49 .65 .58 .49 ----
The results suggest that the QMSB has relatively good reliability in terms of internal consistency. The structure seems suitable for the whole sample, at both Time 2 and 3 and for both males and females. Both the first model, with five types of MSB, and the second model, with one higher-order overall type of MSB and five first-order factors, represent a coherent conceptual and empirically validated framework for assessing student’s MSB in secondary vocational education. The models are almost equivalent in chi-square test and fit indices. The difference is that the second model provides a concrete representation of the covariance between the first-order factors.
The question now arises, whether or not it is useful to distinguish five separate scales, when a simpler model (1-overall factor for MSB) provides an (almost) equally good fit for the data. In the final section this will be investigated.
Competing Models
In order to determine which of the competing models is best suited to provide insight into the data gathered in our longitudinal research project we assessed the relationship between predictive measures of MSB and MSB in the school context as operationalized by the two competing models. In the first model direct effects from predicting variables on the five subscales of MSB were assessed. In the second model only indirect effects through the latent overall factor of MSB are admitted (see Figures 2 and 3). Data from the second data wave were used, because in this sample sufficient boys and girls are represented to take ‘gender’ into account. Three aspects of the school climate that were expected to be related to the different types of MSB were selected, namely School Alienation, Perceived Instructional Support, and Perceived Competitiveness.
School Alienation is expected to have a positive relation with MSB. Various researchers have demonstrated a significant relationship between students’ feelings of disengagement to school and delinquent behaviour (Garnefski, 2000; Cheung, 1997), truancy (Garnefski, 2000), and disruptive behaviour at school (Ryan & Patrick, 2001).
Furthermore from the perspective of motivational self-regulation, it is expected that when personal goals can not be achieved in the school context and school goals are not adopted by the students, negative feelings arise. It is probable that students who report school alienation identify themselves to a lesser extend with school norms and rules and are less reluctant in trespassing these school norms. In this research ‘School Alienation’ is measured by two items: ‘I regularly think of changing my course’, ‘to be honest, I’m looking for a new course’ (α = .90, M = 2.41, SD = .83).
Clarity of instruction and the way teachers motivate their students to engage in the learning activities is also expected to be related to MSB. Students who perceive that their teachers explain the subject matters in a clear way, will show less off-task behaviour (e.g., disturbing the lesson, chatting, provoking others, withdrawing from the learning process). The perception of Instructional Support is measured by four items (e.g., ‘in our school teachers explain the subject matters in such a way that I can really understand it’, α = .86, M = 2.77, SD = .52).
The third measure is the perception of Competitiveness in the school.
Researchers (e.g., Deci, Koestner & Ryan, 2001) have demonstrated that students who perceive a mastery oriented reward structure, are more focused on the task and
less distracted by other personal goals in class, whereas students who perceive a more performance oriented reward structure (i.e., in which the best performing students get most attention), are less focused on the task and more open to off-task behaviour. For example they might be more eager to impress their peers by acting out and breaking the rules. This measure is operationalized by four items (i.e., ‘in this school quick students gain more respect than slow students’, α = .89, M = 2.44, SD = .58).
As far as we know, there is no research available on the specific relationships between these three school climate variables and each of the five types of maladaptive behaviour. Therefore prior to examining the differential effects on the two models, a model was tested in which all predictor variables were allowed to relate to all five outcome variables (i.e., all three school climate variables and all five manifest MSB constructs). Non-significant estimates were then one by one removed. A Bonferroni correction was conducted and the critical p-value was changed form .05 in .01. We dichotomized Delinquent Behaviour because this variable was extremely skewed.
Figure 3 shows that Instructional Support, Competitiveness and School Alienation were all significantly related to the overall variable for MSB, for the whole sample, boys, and girls. However, Figure 2 shows that Instructional support is only related to MSB toward Schoolwork and Rules, Unfriendly Behaviour and Impolite Behaviour (for the whole sample and girls), thus not for Delinquent Behaviour and Withdrawn Behaviour. This means that teacher’s instruction can have an effect on the frequency that students distract other students, not participate in the lesson (MSB toward Schoolwork and Rules), and provoke teachers or students (Unfriendly Behaviour). It also means that teacher’s instruction can not prevent students from taking someone’s belongings, fighting (Delinquent Behaviour), or isolating themselves from other students and not joining conversations (Withdrawn Behaviour). With respect to the relevance of Competitiveness and School Alienation, most remarkable is the difference between the genders; Competitiveness seems important in predicting all types of MSB of girls (except Withdrawn Behaviour), School Alienation seems important in predicting all types of MSB of Boys. Hence, it is important to distinct between boys and girls. Several studies (e.g., Ryan & Patrick, 2001; Aalsma &
Lapsley, 2001; Barnes & Welte, 1986) have shown that boys report more MSB than girls. This is (almost without exception) found for all types of MSB, except for withdrawn, anxious, or depressive behaviour (e.g., Nolen-Hoeksema & Girgus, 1994).
Remarkably, only a small percentage of the variance of Withdrawn Behaviour could be predicted for the whole sample and for boys by school climate variables. This might indicate that this type of maladaptive behaviour is not easily influenced by the school climate, and might be an intrinsic part of the students’ personality structure.
Figure 2: Model 1
Parameter estimates are presented for the whole sample/ boys/ girls; Dashed lines indicate negative paths.
Schoolwork and rules
Delinquent behaviour
Unfriendly behaviour
Withdrawn behaviour
Impolite behaviour Instructional
support
Competitive ness
School alienation
.98/.95/1.0 .93/.93/.94
.94/.92/.97 .93/.91/.95
.93/.96/.93
.28/.18/ns .23/.12/ns
.24/.15/ns .17/29/ns
.29/.20/.16 .15/ns/.22
.13/ns/15
.16/ns/24
.10/ns/.18 -.15/-.09/-.13
-.16/-.12/-.15 -.08/-.ns/-.12
.30/.16/.36 .29/.38/.30
121
Figure 3: Model 2
Parameter estimates are presented for the whole sample/ boys/ girls; Dashed lines indicate negative paths.
Schoolwork and rules
Delinquent behaviour
Withdrawn behaviour
Impolite behaviour Instructional
support
Competitive ness
School alienation
MSB
overall .30/.32/.39
.53/.32/.70 Unfriendly
behaviour
.67/.74/.79 .51/.55/.49
.69/.83/.55
.77/.81/.72 .83/.82/.78
.57/.51/.52 .70/.67/.71
.30/.16/.36 .29/.38/.30
.21/.31/.13 .20/.17/.24 -.23/-.24/.20
ns/ns/ns
.41/.35/.49s
122
For the whole sample school climate variables explained 7% of the variance in MSB toward Schoolwork and Rules, Delinquent Behaviour and Unfriendly Behaviour, 6%
in Impolite Behaviour, and 2% in Withdrawn Behaviour. Furthermore the covariance between school alienation and competence development appeared non-significant.
In general more variance of all types of MSB could be predicted in the sample of boys, except for Delinquent Behaviour.
Table 4 shows the Chi-square statistics and fit-indices. Figure 2 shows the standardized estimates for the whole sample, for boys and for girls. Covariances between the behaviour variables are not shown, in order to maintain a readable picture. Model 2 shows the indirect effects through the latent overall factor for MSB (Figure 3). Comparison of the fit-indices of both models, suggests that the first model (with the direct effects) provides the best fit for the data, both for the whole sample, and for boys and girls separately. This model provides the best opportunities for detailed conclusions and for exploring differences between subgroups. On the basis of Model 2 it can (only) be concluded that all selected predictors are significantly related to MSB for the whole sample and for boys. Findings could be interpreted in more detail on the basis of Model 1. Although the statistics show that the model is an adequate representation of the data, only small amounts of variance was explained.
Nevertheless, our findings suggest that it is important that teachers pay close attention to school climate variables, because they might influence specific types of MSB at school.
Table 4: Chi-square statistics and fit-indices for Structural Models.
Model Sample N X2 Df p-value CFI NFI RMSEA
Model 1 All 852 11,17 4 .025 1.00 1.00 .046
Model 1 Boys 301 14,95 9 .102 1.00 .99 .046
Model 1 Girls 524 10,98 8 .207 1.00 .99 .026
Model 2 All 852 151,52 18 .000 .95 .95 .093
Model 2 Boys 301 98,91 18 .000 .95 .94 .093
Model 2 Girls 524 75,72 18 .000 .96 .94 .078
Discussion
The findings of this study provide support for distinguishing different types of MSB at school. A model that takes these different types of MSB into account provides the best opportunities for a detailed description and understanding of the way school climate affects MSB in respective subgroups. Granted, an overall score for MSB is useful and can be used in global analyses, but a detailed interpretation of the data seems to have additional benefits. We emphasize that the mere observation that all types of MSB are highly correlated and thus seem to co-occur, does not automatically warrant the existence of a ‘problem behaviour trait’, as is suggested by other researchers (i.e., Jessor & Jessor, 1977). In this study we were interested in measuring less obvious types of MSB that according to a majority of the reference group (students and staff) is norm deviant, in addition to more serious forms of antisocial behaviour. Differences in severity of the types of MSB might have decreased the likelihood of finding just one construct. Put more simply, a student who behaves ‘unfriendly’ or ‘impolite’ does not automatically participate in delinquent
acts. Also the fact that the QMSB asks specifically for the frequency of MSB at school, could have enhanced the situation specificity of the results; an adolescent who is behaving badly outside school, does not automatically behave badly in the classroom. The situation-specific nature of the items (i.e. referring to the school context and not to MSB in general) has certainly contributed to the variability of the results, and this is in accordance with the theoretical perspective of motivational self- regulation. We argued that antisocial behaviour is conceptualized as the outcome of an interaction process between students’ goals, values and needs and their perception of the learning environment (Boekaerts & Corno, 2005). Students’
perceptions of aspects of the school climate are unstable, varying under influence of personal circumstances (e.g. mood, private circumstances), time of day, perceived course content, contact with other students and many other conditions. As Boekaerts & Corno (2005) argued, students’ self-regulation processes (i.e., the way they target their cognitions, feelings, and actions in the service of their own goals) change accordingly. The challenge for a school is to create a learning environment that students perceive as favourable for learning. Under such circumstances their attention is optimally drawn to the content of the curriculum and their orientation is toward increasing their competences. Boekaerts’ dual processing self-regulation model specifies that students stay on the growth pathway when they perceive the learning goals set by the teacher are concordant with their own goals. They switch to the well-being pathway and engage in off-task behaviour (e.g., pursuing competing goals, such as having fun by disturbing others, provoking the teacher), when they experience that cues in the environment are a threat to well-being. At such a point students engage in behaviour to prevent loss and harm. It is important that researchers and teachers gain insight into the link between perceptions of school characteristics and cues that trigger MSB. Such insight is important not only because engaging in MSB prevents the students from staying on the growth pathway, but also because MSB by peers creates a negative school climate for fellow students and for the teachers. Hopefully research with the QMSB will help us to obtain these insights.
References
Aalsma, M. C., & Lapsley, D. K. (2001). A typology of adolescent delinquency: Sex differences and implications for treatment. Criminal Behaviour and Mental Health, 11, 173-191.
Achenbach, T. M. (1991). Manual for the Child Behaviour Checklist/4-18 and 1991 profile. Burlington, VT: University of Vermont Department of Psychiatry.
Achenbach, T. M. (1991). Manual for the Youth Self Report. Burlington, VT:
University of Vermont Department of Psychiatry.
Asendorf, J. B. (1990). Beyond social withdrawal: Shyness, unsociability, and peer avoidance. Human Development, 33, 250-259.
Bingman, C.R., & Shope, J.T. (2004). Adolescent problem behaviour and problem driving in young adulthood. Journal of Adolescent Research, 19, 205-223.
Björkqvist, K., Lagerspetz, K. M. J., & Kaukiainen, A. (1992). Do girls manipulate and boys fight? Developmental trends in regard to direct and indirect aggression.
Aggressive Behaviour, 18, 117-127.
Bleichrodt, N., Resing, W. C. M., & Zaal, J. N. (1993). Handleiding voor de Schoolgedrag Beoordelingslijst (SCHOBL-R). [Manual for the Schoolbehaviour Assessment list]. Lisse: Swets & Zeitlinger Publishers.
Blöte, A. W., & Curfs, L. M. G. (1986). Het gebruik van de Conners Teacher Rating Scale in Nederland [The use of the Conners Teacher Rating Scale in The Netherlands]. Nederlands Tijdschrift voor de Psychologie, 41, 226-236.
Boekaerts M. (2005). Self-Regulation: with a focus on the regulation of motivation and effort. In W. Damon & R, Lerner (Series Editors) Handbook of Child Psychology, Vol 4., Child Psychology in Practice (I.E. Sigel & K.A. Renninger, volume editors). New-York: Wiley.
Boekaerts, M., & Corno, L. (2005). Self-regulation in the classroom: A perspective on assessment and intervention. Applied Psychology, 54, 199-231.
Browne, M. W. (1984). Asymptotically distribution-free methods for the analysis of covariance structures. British Journal of Mathematical and Statistical Psychology, 37, 62-83.
Browne, M. W. & Cudek, R. (1992). Alternative ways of assessing model fit.
Sociological Methods & Research, 21, 230-258.
Cheung, Y. W. (1997). Family, school, peer, and media predictors of adolescent deviant behaviour in Hong Kong. Journal of Youth and Adolescence. 26, 569- 596.
Cochran, M., & Bo, I. (1989). The social networks, family involvement, and pro- and antisocial behaviour of adolescent males in Norway. Journal of Youth and Adolescence, 18, 377-398.
Deci, E. L., Koestner, R., & Ryan, R. M. (2001). Extrinsic rewards and intrinsic motivation in education: Reconsidered once again. Review of Educational Research, 71, 1-27.
DiNapoli, P.P. (2003). Guns and dolls: An exploration of violent behaviour in girls.
Advances in Nursing Science, V26 (2), 140-148.
Dowson, M., & McInerney, D.M. (2001). Psychological parameters of students’
social and work avoidance goals: A qualitative investigation. Journal of Educational Psychology, 93, 35-42
Elliot, D. S., & Ageton, S. S. (1980): Reconciling race and class differences in self- reported and official estimates of delinquency. American Sociological Review 45, 95–110.
Farrell, A. D., Kung, E. M., White, K. S., & Valois, R. F. (2000). The structure of self- reported aggression, drug use, and delinquent behaviours, during early adolescence. Journal of Clinical Child Psychology, 29, 282-292.
Farrington, D. P. (1995). The development of offending and antisocial behaviour from childhood: Key findings from the Cambridge study in delinquent development. Journal of Child Psychiatry, 360, 929-964.
Farrington, D. P., Loeber, R., Stouthamer-Loeber, M., & Schmidt, L. (1996). Self- reported delinquency and a combined delinquency seriousness scale based on boys, mothers and teachers: Concurrent and predictive validity for African-Americans and Caucasians. Criminology, 34, 493-517.
Garnefski, N. (2000). Age differences in depressive symptoms, antisocial behaviour, and negative perceptions of family, school, and peers among adolescents.
Journal of the American Academy of Child and Adolescent Psychiatry, 39, 1175-1181.
Haynie, D.L., Nasel, T., Eitel, P., Crump, A.D., Saylor, K., Yu, K., & Simons-Morton, B. (2001). Bullies, Victims, and bully/victims: distinct groups of at-risk youth.
Journal of early adolescence, 21, 29-49.
Hijzen, D., Boekaerts, M., & Vedder, P. (2006). The relationship between students’
goal preferences and the quality of cooperative learning. Scandinavian Journal of Psychology, 47, 1, 9-21.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modelling, 6, 1-55.
Jessor, R. J., & Jessor, A. L. (1977). Problem behaviour and psychosocial development. New York: Academic Press.
Jöreskog, K. G. (2004). Structural equation modelling with ordinal variables using
LISREL. Retrieved 13 January, 2005 from
www.ssicentral.com/lisrel/corner.htm.
Koerhuis, M. J. C., & Boekaerts, M. (2006). Maladaptive social behaviour of adolescent in a school context. Leiden University, The Netherlands:
Dissertation.
Kowalski, R. M. (2001). Behaving badly. Aversive behaviours in interpersonal relationships. Washington, DC: APA.
Lemos, M. S. (1996). Students' and teachers' goals in the classroom. Learning and Instruction, 6, 151-171.
Loeber, R. (1982). The stability of antisocial and delinquent child behaviour: A review. Child Development, 53, 1431-1446.
Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. A. (2001). Sex differences in antisocial behaviour. Conduct disorder, delinquency, and violence in the Dunedin Longitudinal Study. Cambridge: University Press.
Mortimer, J.T., Finch, M., Shanahan, M., & Ryo, S. (1992). Work experience, mental health, and behavioural adjustment in adolescence. Journal of Research on Adolescence, 2, 25-57.
Neuvel, J. (2004). Monitor Sociale Veiligheid in de BVE-sector 2004. Deel 1:
Deelnmers. [Monitor Social Security in secondary vocational education 2004.
Part 1: Students]. ‘s Hertogenbosch: CINOP.
Newcomb, M. D., Maddahian, E., & Bentler, P. M. (1985). Risk factors for drug use among adolescents: concurrent and longitudinal analysis. American journal of public health, 75, 525-531.
