Vous êtes sur la page 1sur 14

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

USING LISREL PROGRAM FOR EMPIRICAL RESEARCH


Olusegun Ayodele Adelodun
Institute of Education, Obafemi Awolowo University, Ile-Ife, Nigeria.

Olushina Olawale Awe


Department Of Mathematics, Obafemi Awolowo University, Ile-Ife, Nigeria.

Abstract:
In the last three decades, the LISREL model, methods and software have become
synonymous with structural equation modeling (SEM) which allows researchers in the social
sciences, management sciences, behavioral sciences, biological sciences, educational sciences
and other fields to empirically assess their theories. These theories are usually formulated as
theoretical models for observed (manifest) and unobservable (latent) variables. When data are
collected for the observed variables of the theoretical model, the LISREL program can be
used to fit the model to the data. In this paper, we highlight the various updated versions and
demonstrate the usability of the LISREL program in empirical research using a new empirical
example of cross relationships between latent constructs and students academic
performance.

Keywords: LISREL Program, SEM, Empirical Modeling, Educational Performance.

1. Introduction
The LISREL methodology development started in 1970, when Karl Joreskog
presented a first LISREL model at a conference. The first generally available LISREL
program (version 3) was published in 1975. The name LISREL is an acronym for Linear
Structural Relations. The qualifier Linear is too restrictive for the current version of the
LISREL program, but the name LISREL has become synonymous with Structural Equation
Modeling or SEM. However, LISREL for Windows is no longer limited to SEM. The latest
LISREL for Windows includes the following statistical applications:

LISREL for structural equation modeling.

PRELIS for data manipulations and basic statistical analyses.


1

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

MULTILEV for hierarchical linear and non-linear modeling.

SURVEYGLIM for generalized linear modeling.

MAPGLIM for generalized linear modeling for multilevel data.

ISSN 1857-8047

In 1986, the first version of PRELIS was published. Its main function is the
computation of the appropriate summary for analysis with the LISREL program. Over time,
its functionality has expanded to include a variety of exploratory tools to give users a better
understanding of their data before they attempt a LISREL model. Its name stands for
preprocessor for LISREL.
SIMPLIS was publicly introduced in 1993 as an alternative syntax for the LISREL
model specification. As the name implied, it meant a tremendous simplification over the
original LISREL syntax, which required some understanding of matrix algebra as well as the
memorization of several Greek characters. Another added capability was the drawing of path
diagrams, with some first attempts to allow the user to interact with the program. Up to that
point, the user communicated with the program in so-called batch mode. Some kind of script,
also called syntax file or command file, had to be written with a text editor, containing
instructions for the program: instructions about where the data file was to be found, how
many variables it contained, the names of the variables, etc. This command file was then
submitted to the actual program for execution.
In 1998, the first truly interactive version of LISREL (version 8.20 for MS Windows)
introduced a dialog-box interface to facilitate the writing of those command files. As much as
possible, typing of syntax was replaced with point-and-click actions in a series of dialog
boxes. From these actions, the program then produced the equivalent command file, in
PRELIS syntax for processing of raw data, or in SIMPLIS or LISREL syntax for the actual
SEM analysis.
The path diagram feature became much more flexible, giving maximum control to the
user, not only to specify or change a model, but also to produce a graphic representation of
the model with publication quality that could be exported as a separate file for subsequent
import in document. Parallel with these developments regarding the way that the user
interacts with the program, numerous statistical features were added, including a multilevel
analysis module, exploratory factor analysis and principal component analysis, imputation for
missing values, bootstrap and Monte Carlo procedures, etc. More details of the other versions
of can be found at http://www.ssicentral.com/lisrel/.

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

The LISREL program makes all these additions available to the user through a choice
of menus and submenus, toolbars, and command buttons. LISREL interface has developed to
the point that this paper becomes necessary for beginners.

2. A Succinct Review of LISREL as a Vital Tool for Research Analysis


Researchers using LISREL as a tool for analysis have increased dramatically over the
past few decades (MacCallum & Austin, 2000). LISREL is a flexible modeling tool for many
multivariate statistics, such as regression analysis, path analysis, factor analysis (e.g., Bentler,
1995; Jreskog & Srbom, 1996), ANOVA to MANCOVA (Bagozzi & Yi, 1999), Canonical
Correlation Analysis (Fan, 1997), and Growth Curve Modeling (Ducan, Ducan, Strycker, Li,
& Alpert, 1999). Also, Raykov (2001) presented a collection of examples showing how
multivariate covariance and mean structure hypotheses could be tested via LISREL. His
article focuses on how to test covariance hypotheses, such as MANOVA, using SEM. We
shall therefore review some work done by researchers with the help of LISREL program.
Adelodun (2008) identified the variables that tend to affect educational performance
among adult learners, and developed structural equation models (SEM) for examining the
relationships between the variables. He also estimated the parameters of the models and
evaluated the formulated models. This was with a view to providing an appropriate
framework for predicting educational performance. The mathematical framework of SEM


used was of the form mx1 Bmxm mx1 mxn nx1 , where mx1 and nx1 are vectors of latent
and manifest dependent variables and independent variables respectively. The corresponding
and for dependent and independent variables
design covariance matrices were

respectively. Data collected from a sample of 2000 students were analyzed using the factor
analysis tool of factor loadings, percentages, F test, 2 test and structural equation model
technique. Seven factors were identified to affect educational performance. These were
parental socio-economic characteristics (0.517), circumstances (0.604), self concept (0.647),
health characteristics (0.666), marital status (0.730), training environment (0.796) and
parenting style (0.817), the figures in parentheses being factor loadings. The study concluded
that structural equation model was capable of predicting educational performance using
appropriate indicators.
The evaluation of fit of a theoretical model to data is a classic problem and practical
concern in structural equation modeling. The limitations of likelihood ratio tests for assessing
model goodness-of-fit has led to the construction of a variety of descriptive fit indexes for
3

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

this purpose. Johnson and Bodner (2007) reviewed these limitations and propose an
alternative testing procedure for confirmatory tetrad tests.
Flouri (2006) investigated the long-term effect of mothers and fathers interest in
childs education at age 10 and childs locus of control and self-esteem at age 10 in
educational attainment at age 26; and also explored if mothers interest and fathers interest in
childs education are linked to childs educational attainment via their effect in increasing
childs self-esteem and internal locus of control.

The study used longitudinal data from

sweeps of the 1970 British Cohort Study (BCS70). The initial sample was those 1,737 men
and 2,033 women with valid data on age 10 self-esteem, locus of control, fathers interest,
mothers interest, and age 26 educational attainment. Of these, 1,326 men 1,578 women were
included in the final analysis.
Bauer (2005) considered an indirect application of the Structural Equation Mixture
Model (SEMM) in which the latent classes are estimated only in the service of more flexibly
modeling the characteristics of the aggregate population as a whole. More specifically, the
SEMM is used to semiparametrically model nonlinear latent variable regression functions.
This approach is first developed analytically and then demonstrated empirically through
analyses of simulated and real data.
In Simons, Dewitte and Lens (2004), two theories in the field of motivation and
achievement, namely the future time perspective theory and goal theory, result in conflicting
recommendations for enhancing students motivation, because of their differential emphasis
on the task at hand and on the future consequences of a task. This study investigates whether
type of instrumentality a learning task evokes has an influence on students motivation, goal
orientation, and study behaviour and ultimately on performance during the first year training
in nursing. The path analysis shows that both variables independently enhance task
orientation, the use of deep level strategies (via task orientation), persistence (via task
orientation), excitement (via task orientation), regular studying (via task orientation) and
performance. Further, approach and avoidance age orientation decreases, and with it surface
level processing, which also increases performance.
In Mullis, Rathge and Mullis (2003), utilizing the National Education Longitudinal
Study of 1988 (NELS: 88), their study examined some of the frequency used indicators of
social capital and resource capital as predictors of academic performance of 24,599 middle
school adolescents (12,111 males and 12,244 females) ranging in age from 13-16 years. 68%
were White, 12.2% were Black, and 12.9% were Hispanic. The participants were drawn from
a stratified national sample of over 1000 public and private schools in the United States. Data
4

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

from NELS: 88 were analyzed using indicators of social capital and resource capital.
Preliminarily modeling indicated the need to separate social capital into two components:
parental networks and student networks. Resource capital, including parent education, parent
income, and educational items in the home, was most predictive of academic performance. In
addition, student reported misbehavior (behavior) in school was included as a mediating
variable. The findings indicated that both indicators of social capital were not strong
contributors to academic performance among adolescents. Resource capital indicators were
found to be stronger contributors to academic performance. The context variable
(hypothesized as a mediator variable) of student misbehavior in school was found to be the
best predictor of academic performance. The results are discussed in relation to social capital
theory and future research of viable predictors of academic performance among adolescents.
The study examined by Verlaan and Schwartzman (2002) links between parents
adjustment and childrens externalizing behaviour problems representative of the general
population. Structural equation modeling was used to examine models that included direct
and indirect pathways. Externalizing behaviour problem in children were strongly related to
parental adjustment difficulties. Maternal antisocial behaviour and marital hostility were
licked to sons and danglers externalism behaviour problems via dysfunctional childrearing
practices. Maternal antisocial behaviour, however, was not fully mediated by dysfunctional
parenting. For father the pattern related to externalizing problems in sons was as those
specified for mothers. The pattern did not appear relevant for daughters. Results are discussed
in terms of the importance of adjustment patterns with regard be the sex composition of the
parent child dyads.
Ganzach (2000) examined the interactions between parents education, cognitive
ability and education expectation in determining educational attainment. The information was
taken from the National Longitudinal survey of youth. Three interactive relationships
examined in this work fall into two general classes of interactive relationships: synergistic
relationships and offsetting relationships. A synergistic relationship implies that high (low)
values of the independent variables result in especially high (low) values of the dependent
variable, higher (lower) than might be expected from a linear combination of the independent
variables. An offsetting relationship implies that it is enough that one of the independent
variables will be high (low) in order for the dependent variable to be high (low).
Dukes, Ullman, and Stein (1995) examined the short-term effects of the D.A.R.E.
(Drug Abuse Resistance Education) program using classroom level data and a Solomon 4group design. In series of 2 group models with latent means, pretest sensitization, maturation,
5

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

and treatment effects were examined for the core concepts (latent variables) targeted by
D.A.R.E.: (1) self-esteem (2) bonds to family, police, and teachers, (3) endorsement of risky
behaviors, and (4) resistance to peer pressure. The maturation analysis of the non-treated
classrooms revealed that self-esteem and bonds to family, police, and teacher significantly
decreased over time. This is opposite to the effects of the D.A.R.E. program. There was
evidence of pre-test sensitization for resistance to peer pressure. Participating in the pre-test
alone led to greater resistance to peer pressure. Regardless of participation in the pre-test,
students who participated in the D.A.R.E. program reported significantly higher self-esteem,
and stronger bonds to family, teachers, and police. They also endorsed fewer risky behaviors
than students who did not participate in D.A.R.E.
Stein, Newcomb, and Bentler (1993) examined the effects of grandparent and parent
drug use on behavior problems in boys and girls aged 2-8. Separate structural equations
models were developed for boys and girls. Six measured child behavior variables served as
DVs: (1) developmental problems, (2) fearfulness, (3) hyperactivity, (4) acting out, (5)
psychosomatic complaints and (6) social problems. The IVs of interest were four latent
variables; (1) grandparent drug use, (2) mothers drug problem, (3) mates drug use, and (4)
dyadic adjustment. They found that boys were more likely than girls to have developmental
and social problems. Drug use among grandparents and mothers was associated with more
problems in boys than in girls. Grandparent drug use directly predicted greater hyperactivity,
more acting out, more psychosomatic complaints, and greater social problems in boys. In
girls, grandparent drug use was associated with more acting out. Maternal drug use also
predicted more fearfulness, hyperactivity, and social problems for boys.
Fullagar, McCoy, and Shull (1992) examined the effects of union socialization
programs on union attitudes and loyalty in a path analysis model. The participants in this
study were 71 apprentices undergoing a union-management training program. Measures of
(1) subjective norms, (2) charismatic leadership, (3) individual consideration, (4) intellectual
stimulation and (5) training satisfaction were employed as IVs to predict (1) union attitudes,
(2) union socialization and (3) union loyalty. Union socialization also served to predict union
attitudes. They found the most important predictors of union loyalty were positive union
attitudes, and training satisfaction. Charismatic leadership and individual consideration both
predicted union socialization. Union socialization predicted positive union attitudes.
Subjective norms, individual consideration, and intellectual stimulation predicted positive
union attitudes.

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

Harlow and Newcomb (1990) tested several confirmatory factor analytic models to
describe the relationships among 25 measured variables related to meaning and satisfaction in
life. The best model consisted of 9 first-order specific factors, 3 second-order primary factors,
and one third-order general factor. Beginning at the top, the third-order factor representing
Meaning and Satisfaction in Life predicted the three second-order factors; (1) Relationship
Satisfaction, (2) Purposeful Living, (3) Work and Health Satisfaction. The Relationship
Satisfaction predicted three specific first-order factors; (1) Peer Relationships, (2) Intimate
Relationships, (3) Family Relationships. The Purposeful Living factor also predicted three
specific first-order factors; (1) Purposeful in Life, (2) Meaninglessness, and (3)
Powerlessness. Finally, the Work and Health Satisfaction factor predicted two specific firstorder factors; (1) Work Satisfactions and (2) Health Satisfaction.
Era, Jokela, Qvarnverg, and Heikkinen (1986) studied pure-tone thresholds and
speech understanding in three different age groups of men: (1) 31-35, (2) 51-55, (3) 71-75.
Cognitive capacity, occupational status, number of symptoms, and noise at work used to
predict hearing threshold. The strongest predictors of hearing threshold at 4,000 Hz in the 3135 year old group were occupational status and number of symptoms. The equation
accounted for 22% of the variance in hearing threshold. Overall, the strongest predictors of
auditory functioning within each age group were occupational status, noise exposure at work,
cognitive capacity, education, and number of disease symptoms.
Hence there is need for researchers in the fields of Science, Technology and Education to be
further familiar with and understand the importance and usability of LISREL program in
empirical research, which this work is aimed at achieving

3.0 Empirical Analysis


The conceptual model considered in this empirical study consists of Parenting Style
with three manifest variables: Permissive, Authoritarian, and Authoritative; and also the
Training Environment with two manifest variables: Boarding facility, and School type. The
indicators for measuring students academic performance are parents socio-economic status,
parents educational status, and aptitude test.
The following three hypotheses are proposed:
H01: There is no significant relationship between Parenting styles and Students academic
performance.
H02: There is no significant relationship between Training environment and Students
academic performance.
7

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

H03: There is no significant relationship between Parenting styles and Training environment
The research design used for the study is survey research. One Hundred and Nineteen (119)
students of the Faculty of Education, Obafemi Awolowo University, Ile-Ife, were selected
randomly from the 7 departments (Special Education and Curriculum studies; Institute of
Education; Educational Foundation and Counselling; Department of Educational Technology;
Department of Continuing Education; Physical and Health Education; and Educational
Administration and Planning). The selected study sample, from these departments, are
grouped into Sciences and Social Sciences based on their programme of study.
The conceptual model for this study is to examine the cross relationships between
latent constructs and students academic performance.

Parents socioeconomic
status

Permissive

Acad per

Aptitude test

H1
Authuorita
rian

H2
Parenting

Boarding
facility

H3
School type
Training
Env.

Authorita
tive

Parents
Educational
Status

Figure 1: Conceptual Model for Parenting styles and Training environment on Academic
performance

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

Table 1. Profile of the Students


Frequency Percentage
Gender

Male

54

45.4

Female

65

54.6

Sciences

59

49.6

Social Sciences

60

50.4

Parents Socio-

Wealthy

76

63.9

economic status

Not wealthy

43

36.1

Parents Educational

Father alone

12

10.1

Status

Mother alone

2.5

Both parents

90

75.6

None of them

14

11.8

Programme

The research instrument adopted is the standardized Parental Authority Questionnaire


(PAQ) with its constructs identification. The data collected was analyzed using frequency
count, percentage score, descriptively, while a structural equation modeling technique was
used to test the model, inferentially. The LISREL 8.80 program was employed. The
covariance matrix of the variables in the model are given in the Appendix of this paper.

3.1 Analysis of the Structural Model


The parameters of the model are estimated with the Maximum Likelihood (ML)
procedure. The goodness-of-fit was evaluated with indicators: Chi-square/degrees of freedom
(x2/df); comparative fit index (CFI); non-normed fit index (NNFI); goodness-of-fit index
(GFI) criteria, among others.
Hypothesis testing was performed on the model. The observed (manifest) variables or
indicators (Permissiveness, Authoritarian, Authoritative, Boarding facilities, School type,
Parents Educational Status, Parents Socio-economic Status, and Aptitude test) were used to
predict the latent variables (Parenting styles, Training environment, and Academic
performance). The Structural Equation Modeling was obtained by processing the data in the
instrument. Table 2 reveals the observed variables as against the corresponding latent
variables.

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

Table 2 Associated Latent Variables versus Manifest/Observed Variables.


Latent Variable

Manifest Variable

Abbrevation

Permissive

Permiss

Authoritarian

Authrian

Authoritative

Authtive

Boarding facility

Brdfac

Kind of School

Schtype

Parents Educational Status

Pes

Academic

Parents Socio-economic

pses

Performance

Status.

Parenting Styles

Training Environment

Aptitude Test

apt

The Manifest variable pses; pes; and apt, were used as predictors of the Latent
variable academic performance. These variables were calculated using the relevant items as
a result of the factor analysis carried out on the items (Adelodun and Obilade, 2011), in the
data collection tool, which tend to explain the effect of Parenting styles and Training
environment on the students academic perfofmance.
Seven fit indexes which are commonly used in the literature i.e. x 2/df; GFI; AGFI;
NNFI; CFI; RMSR; RMSEA, were employed to test the model fit. The commonly used
measures of model fit, based on results from an analysis of the Structural Model, are
summarized in Table 3. In practice, Chi-square/degrees of freedom (x2/df) less than 3; GFI
greater than 0.9; NNFI greater than 0.9; CFI greater than 0.9; an AGFI greater than 0.8;
SRMSR less than 0.1; and RMSEA less than 0.06 or 0.08, are considered indicators of good
fit.

Table 3 Summary Statistics of Model Fit.


Fit index
Chi-square/degrees of freedom
GFI
AGFI
NNFI
CFI
SRMSR
RMSEA

Recommended value
3.00
0.90
0.80
0.90
0.90
0.10
0.06 or 0.08

Observed value
1.18
0.95
0.91
0.69
0.79
0.07
0.04

10

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

GFI = Goodness-of-Fit Index


CFI = Comparative Fit Index
AGFI = Adjusted Goodness-of-Fit Index
SRMSR = Standardized Root Mean Square
Residual
NNFI = Non-normed Fit Index
RMSEA = Root Mean Square Error of
Approximation

Chi-square=22.43, df=19, p-value=0.26045, RMSEA=0.039


Figure 2: LISREL test of Research Model
Graphical presentation of the results is shown in Figure 2. Figure 2 shows the standardized
LISREL path coefficients with their respective significance levels.

4. Main Results
From the members of the research group, 54 (45.4%) students are male, and 65
(54.6%) students are female. 59 (49.6%) students are Sciences while 60 (50.4%) students are
Social Sciences. Majority of the respondents, 76 (63.9%) students are from wealthy families
but 43 (36.1%) students are not. Also, 12 (10.1%) students have only their father educated, 3
(2.5%) students have only their mother educated, 90 (75.6%) students have both of their
parents educated, and 14 (11.8%) students have none of their parents educated.
As seen in Table 3, all good-of-fit statistics are in the acceptable ranges except for
NNFI and CFI which are close enough (0.69 and 0.79 respectively) to the recommended
value 0.90.

11

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

Figure 2 illustrates the significant structural relationships among the study variables.
Hypothesis 1 and 2 are postulated that Parenting styles and Training environment have
significant effect on Students academic performance. The direct path Parenting
styleAcademic performance is significant though reversed since the regression co-efficient
() is -0.09 with t = -1.17, and p<0.05. Therefore the hypothesis is supported, which means
that the style of parenting significantly has a direct effect on students academic performance.
Moreover, the direct path Training environment Academic performance is significant since
is 0.17 with t = -1.42, and p<0.05. The third hypothesis is also accepted because the direct
path Parenting styleTraining environment is significant since is 0.50 with t = 0.96, and
p<0.05.

5. Conclusion/Summary
This paper attempted to explain the concept of SEM using the LISREL (Linear
Structural Equations) approach: its major purpose, application, types of models, steps
involved in formulation and testing of models.
We reveal that the LISREL 8.80 for Windows (Jreskog & Srbom, 2006) is a
suitable Windows software product for structural equation modeling, multilevel structural
equation modeling, generalized linear modeling, multilevel linear modeling, multilevel
generalized linear modeling, and multilevel nonlinear modeling, among others.

References:
Adelodun, O.A. (2008): Application of Structural Equation Modeling to Latent Constructs for
Evaluating Educational Performance of Adult Students. Unpublished Ph.D. (Mathematics)
Thesis, Obafemi Awolowo University, Ile- Ife. Nigeria.
Adelodun, O.A. & Obilade, T.O. (2011). Identification of Factors Affecting Educational
Performance of Nigerian Adult Learners: A preliminary Study. African Research Review:
An International Multi-Discipline Journal, Ethiopia. Vol. 5(2), S/No. 19 pp. 225- 232.
Bagozzi, R.P. & Yi, Y. (1999). On the use of structural equation models in experimental
designs. Journal of Marketing Research, 26, 271-285.
Bauer, D.J. (2005). A Semiparametric Approach to Modeling Nonlinear Relations Among
Latent Variables. Structural Equation Modeling, 12(4), 513535, Lawrence Erlbaum
Associates, Inc.
Bentler, P.M. (1995). EQS structural equations program manual. Encino, CA:
MultivariateSoftware.
12

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

Ducan, T.E., Ducan, S.C., Strycker, L.A., Li, F. & Alpert, A. (1999). An introduction to
latent variable growth curve modeling. Mahwah, NJ: LEA.
Dukes, R.L., Ullman, J.B. & Stein, J.A. (1995) An evaluation of D.A.R.E. (Drug Abuse
Resistance Education) using a Solomon Four-Group Design with latent variables. Evaluation
Review.
Era, P., Jokela, J., Qvarnverg, Y. & Heikkinen, E. (1986). Pure-tone threshold, speech
understanding and their correlates in samples of men of different ages, Audiology, 25, 338352.
Fan, X. (1997). Canonical correlation analysis and structural equation modeling: What do
they have in common? Structural Equation Modeling, 4, 65-79.
Flouri, E. (2006). Parent interest in childrens education, childrens self-esteem and locus of
control, and later educational attainment: twenty-six year follow-up of the 1970 British birth
cohort. British Journal of Educational Psychology, 76, 41-55.
Fullagar, C., McCoy, D. & Shull, C. (1992). The socialization of union loyalty. Journal of
Organization Behavior, 13, 13-26.
Ganzach, H.K. (2000). Does family structure really influence educational attainment? Social
Science Research, 28, 331-357.
Harlow, L.L. & Newcomb, M.D. (1990) Towards a general hierarchical model of meaning
and satisfaction in life. Multivariate Behavioral Research, 25, 387-405.
Http://www.ssicentral.com/lisrel/
Johnson, T.R. & Bodner, T.E. (2007). Some properties of likelihood ratio and nested tetrad
tests comparing factor models with different numbers of factors. Paper presented at the
International Meeting of the Psychometric Society, Tilburg, The Netherlands.
Jreskog, K.G. & Srbom, D. (1996). LISREL 8: Users reference guide. Chicago: Scientific
Software International.
------------ (2006). LISREL 8.80 for Windows [Computer Software].Lincolnwood, IL:
Scientific Software International, Inc.
MacCallum, R.C. & Austin, J.T. (2000). Applications of structural equation modeling in
psychological research. Annual Review of Psychology, 51, 201-226.
Mullis, R.L., Rathge, R. & Mullis, A.K. (2003) Predictors of academic performance during
early adolescence: a contextual view. International Journal of Behavioral Development,
27(6):541-548.
Raykov, T. (2001). Testing multivariate covariance structure and means hypotheses via
structural equation modeling. Structural Equation Modeling, 8, 224-256.
13

Transnational Journal of Science and Technology

August 2013 edition vol.3, No.8

ISSN 1857-8047

Simons, J., Dewitte, S. & Lens, W. (2004). The role of different types of instrumentality in
motivation, study strategies, and performance: know why you learn, so youll know what you
learn! British Journal of Educational Psychology, 74, 343-360.
Stein, J.A., Newcomb, M.D. & Bentler, P.M. (1993) Differential effects of parent and
grandparent drug use on behavior problems of male and female children. Development
Psychology, 29, 31-43.
Verlaan, P. & Schwartzman, A.E. (2002). Mothers and fathers parental adjustment: links to
externalizing behaviour problems in sons and daughters. International Journal of Behavioral
Development, 26(3), 214-224.

Appendix
Covariance Matrix of the Manifest Variables
Pses

Apt

Brdfac

Schtype

Pes

Permiss

Authrian Authtive

Pses

0.23

Apt

- 0.05

-3.06

Brdfac

0.01

-0.14

0.30

Schtype

0.09

-0.13

0.06

0.79

Pes

- 0.01

-0.09

- 0.01

0.16

0.47

Permiss

0.14

-1.39

- 0.26

1.53

- 0.02

35.91

Authrian

- 0.26

-1.29

- 0.06

- 0.40

- 0.06

2.68

75.40

Authtive

- 0.25

0.21

- 0.02

- 0.02

- 0.28

3.33

8.48

14.76

14