International Journal of Information and Education Technology, Vol. 3, No.

2, April 2013

An Empirical Analysis of the Antecedents and

Performance Consequences of Using the Moodle Platform
Chun-Che Huang, Yu-Min Wang, Tsin-Wei Wu, and Pei-An Wang


internship learning? (2) The teaching group
Abstract—With the rapid development of the Internet, quasi-experimental research method is divided into the
constructing the e-learning system has been an integral part of traditional control group, a group of online teaching group
many educational institutions. This study uses the Moodle and mixed teaching group. With the experiment, the
system to make up for the shortcomings of learning in a large
class, and find out the difference of learning performance and
effectiveness of learning of different groups is different?
acceptance between traditional learning and digital learning. Up to date, researchers are confronted with a choice among
The study is experimented with the students in the Calculus numerous models to "choose" constructs, or choose a
course. There are two part of research design in the study. The "favored model" and discard alternative models. Therefore,
first part corresponds to learning performance, which uses the Venkatesh [3] developed an integrated management theory,
quasi-experimental method to compare the difference of the Unified Theory of Acceptance and Use of Technology
experimental group, and the control group. The second part
refers to technology acceptance, where the unified theory of
(UTAUT). This theory aims at understanding user behavior
acceptance use of technology (UTAUT) is used. With two intention construct; and furthermore enhances the implication
dimensions: Outcome Expectation and Attainment Value, a of user behavior. Numerous studies has applied UTAUT in
new hybrid technology acceptance model is proposed to different domains and proven more than 70% of studies with
investigate the students’ intentions to use Moodle systems. Data effective experiment results. UTAUT has been applied in
Analysis of this study shows a great promising in providing new e-learning, for example Van Raaij and Schepers [4], Keller,
academic research evidence in e-learning teaching.
Hrastinski and Carlsson [5], and Chiu and Wang [6].
Index Terms—E-Learning, moodle system, UTAUT model, However these studies do not consider the collaborative
learning performance. learning on the Internet over the network learning platform,
e.g., using Moodle.
Moodle is an Open Source Course Management System
I. INTRODUCTION (CMS) which is provided freely. Moodle is programmed in
With the rapid popularization and development of the PHP by Martin Dougiamas, Australia. He developed it as a
Internet, construction of digital learning system is an integral tool for his dissertation which was on a Socio-constructivist
part of many educational institutions. Many researchers approach to learning (Official Website：http://moodle.org/).
pointed out that learning through digital learning platform is The word Moodle was originally an acronym for Modular
helpful for students learning [1], [2]. Object-Oriented Dynamic Learning Environment. It can be
In a large class, students are often up to more than two installed on any personal computer, educational institution
hundred. It often can’t take into account the individual webserver or website, and one also can modify it in what he
differences of each student. We can find the traditional needs under the GNU General Public License. Mainly
teaching methods lacks of effective learning interactive effective is help educators or institutions to serve the majority
learning effectiveness. However, by collaborative learning on of the online learning community. It offers a variety of
the Internet over the network learning platform, students can modular, powerful and flexibility in the use of the teaching
learn together to enhance the interaction between learners and learning environment, stressed WYSIWYG facilitate
learning. This study will design a study to understand the teachers learning activities edit interface [7], [8].
digital learning methods applied to the traditional large
teaching class to enhance learners' learning effectiveness and
understand the impact of learner, an important factor for the II. RESEARCH METHOD
acceptance of digital learning technology. In this study, understanding of the students using the
In this study, to promote and enhance the effectiveness of Moodle e-learning system calculus internship course learning
learning, students use Moodle learning system in the outcomes and technology acceptance factors in the design of
internship courses. The study also desire to understand the the study is divided into two parts: use experimental research
student's acceptance of Moodle system, as well as to explore for learning outcomes, experimental packet comparing the
whether improved learning outcomes for students to change experimental group and control group. The digital learning
the mode of learning. The main research purposes of this technology acceptance is based on the UTAUT as the
study are as follows: (1) What is the important factors that theoretical basis.
affect students' intention when they use Moodle for online
A. Research Design and Hypothesis
In order to understand learning differences of the different
Manuscript received October 25, 2012; revised January 13, 2013.
The authors are with Nation Chi Nan University, Taiwan 545, R.O.C. learning groups used in calculus internship course, this part
(e-mail: ymwang@ncnu.edu.tw). take a quasi-experimental design to experiment. The

DOI: 10.7763/IJIET.2013.V3.267 217

 International Journal of Information and Education Technology, Vol. 3, No. 2, April 2013

experimental subjects are undergraduate 100 students. Total A. ANCOVA

experimental groups were divided into three groups, divided This study carries out the analysis of the effectiveness of
into the control group, experimental group 1 and learning, control group traditional group, based on the
experimental group 2. The control group of students attend traditional way of teaching face-to-face. The test sample was
internship training course in classroom same as before, in taken from the first and second semesters have to attend
way of traditional face-to-face. The experimental group1 freshman calculus internship training, a total of 256 people.
students take the traditional class training, as well participate Experimental group 1 is online group that uses Moodle online
in online internship course. The experimental group 2 teaching. The effective sample size is 63 students. The
students only participate in online internship. Experiments experimental group 2 is a mixed group that also participates
conducted cycle a total of nine weeks. Pretest is based on the in traditional face-to-face courses as well to participate in
grades in the first semester calculus. Posttest is based on the online Moodle teaching. The effective sample size is 41
midterm results. The different modes of teaching should have students. Each group of raw pretest and posttest scores and
different learning effects on students' learning effectiveness. the number of each group number are shown in Table I.
Therefore, this study assumes that the three groups of
students in the effectiveness of learning will vary in different TABLE I: TEST RESULTS
teaching model. Pretest Posttest average
Group average (Dependent N
B. Research Design of Technology Acceptance (covariance) variables)
Traditional group 74.74 62.80 256
This study understands the technology acceptance in Mixed group 49.26 48.80 41
accordance with the Unified Theory of Acceptance and Use Online group 66.06 52.61 63
of Technology (UTAUT) [3], as the theoretical basis to Total 63.35 59.43 360
modify. The main purpose is in the application of the Unified
Theory of Acceptance and Use of Technology (UTAUT) Homogeneity of with-in regression coefficient test is the
affecting students using the Moodle e-learning system factors. experimental treatment group, based on the covariates (X) to
In addition, according to the relevant literature that the results predict the dependent variable (Y) resulting regression
expected and achievements value will affect the intention to coefficient (slope) to be equal, that each of the regression line
use, so we join these two dimensions to UTAUT model to to be parallel to each other before ANCOVA compare. Table
analyze whether also factors that affect students' use of digital II shows the SPSS 12.0 for regression coefficients within the
learning system (Fig. 1) group of this study sample homogeneity test table.

TABLE II: HOMOGENEITY OF REGRESSION COEFFICIENT TEST

Type III Sun Mean
Source of Squares df Square F Sig.
Corrected
71626.336(a) 5 14325.267 80.533 .000
Model
Intercept 668.069 1 668.069 3.756 .053
Group 1225.680 2 612.840 3.445 .033
The total score
36345.167 1 36345.167 204.323 .000
of the last term
Group* The
total score of 734.993 2 367.497 2.06 .128
the last term
Error 62969.927 354 177.881
Total 1406113.000 360

Fig. 1. Reseach framework. a. R Squared = .532 (Adjusted R Squared = .526)

C. Data Analysis Method Homogeneity test results of the regression coefficients in

Table II group (group * the total score of the last term) F
Structural equation modeling analysis is mainly used to value is 2.066; Significant p-value = 0.128> 0.05, below the
analyze the structure of this study established UTAUT study level of significance, to accept the null hypothesis, said three
module. The main analysis method using the Partial Least sets of regression line slope are the same. It means the
Square (PLS) as a research model analysis tools with the relationship between the covariates (total score of the First
software Smart PLS 2.0. Term) and the dependent variable (this semester midterm
grades) will not differ from the independent variable entry
level of the different. It is comply with the assumed of
III. DATA ANALYSIS AND DISCUSSION homogeneity of with-in regression coefficient, that we can
The research data analysis is mainly divided into the continue to do ANCOVA.
questionnaire test data analysis and learning effectiveness 1) Analysis of covariance
analysis. Students take exam in the same day in the classroom. Table III is an ANCOVA test summary table. Exclude the
Total of 105 questionnaires are received. Deduct an invalid impact of pre-test scores (covariates) on the post-test scores
questionnaires, a total of 104 effective samples were used in (the dependent variable), the effects of the independent
the descriptive statistical analysis, confirmatory factor variables on the dependent variables test of F-value = 6.141,
analysis (reliability and validity) and structural equation significant p-value = 0.002 <0.05, achieve the significant
modeling analysis. level and experimental treatment effect is significant. It

218
 International Journal of Information and Education Technology, Vol. 3, No. 2, April 2013

means the post-test scores of the subjects will vary because of Yi [13] proposed confirmatory factor analysis evaluation
the different ways of teaching methods. criteria and Gefen, Straub and Boudreau [14] goodness-of-fit
indicators recommended data to assess. Assess standard
TABLE III: ANCOVA TEST SUMMARY TABLE includes: (a) the factor loadings of the indicators respective
Type III
Sun of Mean
fields significant; (b) The composite reliability of various
Source Squares df Square F Sig. dimensions is higher than 0.7; (c) Average Variance
Corrected
70891.343(a) 3 23630.448 132.053 .000
Extracted (AVE) is higher than 0.5 but we can accept 0.4.
Model
Intercept 261.485 1 261.485 1.461 .228
Because Fornell and Larcker said that if AVE is less than 0.5,
The total score
60417.997 1 60417.997 337.632 .000
but composite reliability is higher than 0.6, the convergent
of the last term validity of the construct is still adequate [15]. We used PLS
Group 2197.904 2 1098.952 6.141 .002
Error 63704.921 356 178.946 various aspects of data analysis to estimate parameters to
Total 1406113.000 360 calculate the dimensions of composite reliability and average
Corrected Total 134596.264 359
extracted variance.
a R Squared = .527 (Adjusted R Squared = .523) The reliability analysis and convergent validity analysis is
obtained. This study various aspects of internal consistency
TABLE IV: MULTIPLE AFTERWARDS PAIRWISE COMPARISON reliability Cronbach's α values range between 0.719 and
95% Confidence 0.924, all of this are higher than the reliability standard 0.7.
Mean Interval
Std.
(I) Group (J) Group Difference
Error
Sig.(a)
Lower Upper
To said consistency reliability of the dimensions within reach
(I-J) significance level that the four dimensions of the framework
Bound Bound
Mixed Tradition
6.584(*) 2.514 .009 1.640 11.527 of this study has some reliability. In convergent validity, in
group group
addition to this study the factor loadings of the items to help
Online
9.756(*) 2.784 .001 4.281 15.232 condition (FC1) data is 0.616, the rest of the items are to
group
Tradition Mixed
-6.584(*) 2.514 .009 -11.527 -1.640 reach the level of significance 0.7 threshold; Each
group group
Online
dimensions composite reliability (CR) are between 0.838 and
3.173 1.920 .099 -0.603 6.948
group 0.946, higher than the standard 0.7; Each dimensions
Online Mixed Average Variance Extracted (AVE) are between 0.565 and
-9.756(*) 2.784 .001 -15.232 -4.281
group group
Tradition 0.815, higher than the standard 0.5. Comply with the
-3.173 1.920 .099 -6.948 .603
group aforementioned scholars convergent validity of each
*. The mean difference is significant at the .05 level dimension inspection standards, the various aspects of this
study have convergent validity.
On the Table IV after pairwise comparison table can be
obtained multiple comparison of covariates is adjusted 3) Discriminant validity analysis
averages of the qualitative differences between the groups In this study, based on Gaski and Nevin test the
compared by multiple post hoc comparison table can be discriminant validity of the two criteria: (1). Correlation
found: coefficient between the two dimensions of less than 1; (2).
 Mixed results are significant in the traditional group and The correlation coefficient of the two dimensions less than
the online group. the individual Cronbach's α reliability coefficient, it means
 Online Group results with traditional group results are not two dimensions have discriminant validity [16]. The other
significantly different. based on the practice of Fornell and Larcker, the test
discriminant validity criteria: (3) The correlation coefficient
B. Measurement Model of the two dimensions less than the square root of the AVE,
1) Reliability analysis and validity analysis said two dimensions has discriminant validity [15], [16]. This
Into the social and behavioral science research, reliability study using the PLS correlation coefficient matrix analysis of
assessment in general can be divided into four indicators, each measure variables. In this study the correlation
namely test-retest reliability, alternative-form reliability, coefficient of the two dimensions are less than 1, in line with
split-half reliability and internal consistency reliability. the standards set by the aforementioned scholars; and the
Regardless of test-retest reliability, alternative-form correlation coefficient of the two different dimensions are
reliability and split-half reliability, all can be called internal smaller than the dimensions of Cronbach's α reliability
consistency reliability. And Cronbach's α internal coefficient, conform to those of the second criterion. Said the
consistency reliability can adopt the most widely reliability various aspects of this study is in line with Gaski and Nevin
indicators [9], [10]. Nunnally [11] has indicated 0.7 to be an proposed test discriminant validity of the two criteria, and
acceptable reliability coefficient. discriminant validity [16].
In the validity Analysis part of this study, confirmatory 4) Structural model
factor analysis of the Partial Least Squares (PLS) construct
When completed reliability and validity testing in PLS
measurement model fit test, to test each construct adequate
measurement model, next is the path for the PLS model
convergent validity and discriminant validity. The following
coefficient test and predict estimated structural model
sequence analysis of convergent validity and discriminant
analysis. In model path coefficient test part to test the
validity.
relationship between the dimensions of research, analysis of
2) Convergent validity analysis whether the path coefficients significant to the study
In this study, based on Anderson and Gerbing [12] hypothesis test. PLS method does not default allocation,
suggested convergent validity analysis criteria, Bagozzi and therefore do not need to test whether the information is in line

219
 International Journal of Information and Education Technology, Vol. 3, No. 2, April 2013

with the assumption of a normal distribution; On the other analysis shown to summarize instructions. En route the path
hand PLS path coefficient significant p-value of the test, so in of the data path for the value of T, T value of 1.96 indicates
order to estimate whether the path coefficients significantly, that the path independent variables to the dependent variable
Felsenstein [17] recommend using BootStrap method. That is reached a significant level and its route to the solid line
the use of the t-value to estimate the p-value, to test the indicates; T value less than significant standard that route at
significance of coefficient the carried out to determine the dotted line. Fig. 2 can also be seen that the degree of
hypothesis whether was established [18]. In the present study, behavioral intentions dimensions explained variance reached
the path coefficient test is test method BootStrap using 76% (R square = 0.7592), mean this study has practical
SmartPLS software default; while the predictive power of the interpretation value.
model is estimated to determine the use of R-squared.
5) Path coefficient test
The path coefficients represent the strength and direction
of the relationship between the variables. When the path
coefficient is positive, indicating a positive influence; when
is a negative value as compared to the negative influence.
Causal model and observed variables and latent variables do
hypothesis testing, after the test, we can get the relevant data,
in order to determine whether the path has a significant, and
to estimate the assumption results whether established.
6) R-squared value of the model predict
R-squared value is the exogenous variables endogenous
variables can explain the variance in the percentage
representatives of the predictive power of the research model,
Fig. 2. Research results.
its values ranging between 0 and 1. When the larger the value,
the better the explanatory power of the model. In general, C. Result and Discussion
R-squared value> 0.67 with practical value, the R-squared 1) In this study, behavioral intentions dimensions R
value represents a moderate explanatory power between 0.33 squared value of 0.7592, the endogenous latent variables
and 0.66, R-squared value between 0.19 and 0.32 is weak R-squared value> 0.67 as having practical value,
explanatory power [19]. R-squared = 0.33 to 0.66, moderate explanatory power,
After confirm each facet has a certain degree of validity R-squared = 0.19 to 0.32, said explain weak capacity and
and reliability, the next that further research model 76% of the explained variance of the degree of
established hypothesis test, analysis how the variables of behavioral intentions dimensions of this study, has a
"Outcome Expectation", "Attainment Value", "Performance practical value.
Expectancy "," Easy-to-use Expectations ", "Social Influence 2) This study architecture Path “Performance Expectancy
", and "Help conditions" to work together to affect the " → Behavioral intentions” the path coefficients (O) is
Behavioral intentions". The following data is computed for
0.336, the path coefficient (M) is 0.326, the T value after
smart PLS statistical software to collate path coefficient T
coefficient test is 4.468. Said research model proven in
values and the corresponding hypothesis and test results of
the significance level α <0.05 standard, the
this study sorted out the test results table. The testing results
performance expectancy has significant positive
“established” is presented in Table VI.
influence on behavioral intentions, H3a hypothesis of
TABLE VI: THE MODEL PATH COEFFICIENT AND HYPOTHESIS VALIDATION this study is established.
(R-SQUARED = 0.7592) 3) This study architecture Path “ Social Influence →
The relationship between Path
T-value Hypothesis Behavioral intentions ” the path coefficients (O) is
dimensions coefficients
0.186, the path coefficient (M) is 0.183, the T value after
Performance Expectancy
→Behavioral intentions
0.336 4.468 H3a coefficient test is 2.817. Said research model proven in
Social Influence → the significance level α <0.05 standard, the social
0.186 2.817 H5a
Behavioral intentions influence has significant positive influence on
Help conditions → behavioral intentions, H5a hypothesis of this study is
0.160 2.005 H6a
Behavioral intentions established.
Gender X Performance
4) This study architecture Path “ Help conditions →
Expectancy →Behavioral -0.235 2.939 H3b
intentions Behavioral intentions ” the path coefficients (O) is
Remark : T values> 1.96 reached * significant；T values> 2.58 reached 0.160, the path coefficient (M) is 0.150, the T value after
** significant；T values> 3.29 reached ***significant coefficient test is 2.005. Said research model proven in
the significance level α <0.05 standard, the help
From the above Table VI that this study architecture road conditions has significant positive influence on
map through structural equation modeling analysis to take behavioral intentions, H6 hypothesis of this study is
Partial Least Square (PLS) analysis can be obtained established.
following Fig. 2 study architecture structure mode path 5) This study architecture Path “Gender X Performance

220
 International Journal of Information and Education Technology, Vol. 3, No. 2, April 2013

Expectancy → Behavioral intentions ” the path [8] W. H. Rice, Moodle e-learning course development, Packt publishing.
[9] J. R. A. Santos, “Cronbach’s alpha: A tool for assessing the reliability
coefficients (O) is -0.235, the path coefficient (M) is of scales,” Journal of extension, no. 2, 1999.
-0.214, the T value after coefficient test is 2.939. Said [10] P. E. Shrout and J. L. Fleiss, “Intraclass correlations: uses in assessing
research model proven in the significance level α <0.05 rater reliability,” Psychological bulletin, vol. 420, no. 2, 1979.
[11] J. Nunnally, Psychometric methods, New York: McGraw, 1978.
standard, the study regulation variables "gender" for the [12] J. C. Anderson and D. W. Gerbing, “Structural equation modeling in
performance expectancy to behavioral intentions path practice: A review and recommended two-step approach,”
has significant positive influence , H3b hypothesis of Psychological bulletin, vol. 411, no. 3, 1988.
[13] R. P. Bagozzi and Y. Yi, “On the evaluation of structural equation
this study is established. models,” Journal of the academy of marketing science, no. 1, pp. 74-94,
6) In this study, H1a, H1b, H2a, H2b , H4a, H4b, and H5b 1988.
hypothesis of this study are not established based on the [14] D. Gefen, D. W. Straub, and M. C. Boudreau, Structural equation
modeling and regression: Guidelines for research practice, 2000.
path coefficients O, the path coefficient M, the T value, [15] C. Fornell and D. F. Larcker, “Evaluating structural equation models
where the significance level α are all less than 0.05 with unobservable variables and measurement error,” Journal of
marketing research, pp. 39-50, 1981.
[16] J. F. Gaski and J. R. Nevin, “The differential effects of exercised and
unexercised power sources in a marketing channel,” Journal of
IV. CONCLUSIONS marketing research, pp. 130-142, 1985.
[17] J. Felsenstein, “Confidence limits on phylogenies: an approach using
We get the following main conclusions from two aspect, the bootstrap,” Evolution, vol. 39, no. 4, pp. 783-791, July 1985.
Technology acceptance behavior and learning achievement. [18] L. K. Molinari, R. Abratt, and P. Dion, “Satisfaction, quality and value
Study confirmed in Technology acceptance behavior: (1) The and effects on repurchase and positive word-of-mouth behavioral
intentions in a B2B services context,” Journal of Services Marketing,
performance expected is positive significant impact on no. 5, pp. 363-373, 2008.
behavioral intentions. (2) Social Influence is positively [19] W. W. Chin, B. L. Marcolin, and P. R. Newsted, “A partial least
significant impact on behavioral intentions. (3) Help squares latent variable modeling approach for measuring interaction
condition is positive significant impact on behavioral effects: Results from a Monte Carlo simulation study and an
electronic-mail emotion/adoption study,” Information Systems
intentions. (4) Adjustment variables gender performance Research, no. 2, pp. 189-217, 2003.
expectations to behavioral intentions path forward significant
impact. (5) The degree of influence behavioral intentions Chun-Che Huang received his Ph.D. degree in
Industrial Engineering from the University of Iowa,
dimensions explained variance of 76%, with the practical Iowa City, and his M.S. degree in Operations Research
interpretation. from Columbia University, New York, NY. He is a
Study in learning achievement confirmed: (1) Mixed group Professor in the Department of Information Engineering,
National Chi Nan University, Taiwan and directs the
of students learning achievement than traditional groups and Laboratory of Intelligent Systems and Knowledge
online groups of students learning achievement. (2) Students Management (the ISKM Lab.). He is interested in
attending classes online group and traditional group teaching intelligent systems, knowledge management, and data
mining. He has published research papers in journals sponsored by various
methods, there was no significant difference in learning societies.
achievement between the two groups. Learned from the
results, performance expected, Social Influence and Help Yu-Min Wang is an associate professor in Information
Management at National Chi Nan University in Taiwan.
condition is helpful for students to use Moodle system. In He received his Ph.D. degree in Information
terms of learning achievement, the mixed group of students Management at National Sun Yat-Sen University in
learns achievement better than the online group and Taiwan. His research interests include enterprise
traditional group, while no significant difference in the information system, information system adoption,
electronic commerce, e-learning, and knowledge
effectiveness of online group and traditional group learning. management.

REFERENCES
[1] J. Gil-Flores, J. J. Torres-Gordillo, and V.-H. Perera-Rodrí guez, “The Tsin-Wei Wu is a graduate student in national Chi Nan
role of online reader experience in explaining students’ performance in University. He is interested in e-learning and data
digital reading,” Computers & Education, pp. 653-660, March 2012. mining.
[2] R. C. Clark and E. Mayer, E-learning and the science of instruction:
Proven guidelines for consumers and designers of multimedia learning:
Pfeiffer, 2011
[3] V. Venkatesh, M. G. Morris, G. B. Davis, and F. D. Davis, “User
acceptance of information technology: Toward a unified view,” Mis
Quarterly, pp. 425-478, 2003.
[4] E. M. Van Raaij and J. J. L. Schepers, “The acceptance and use of a
virtual learning environment in China,” Computers & Education, vol.
50, no. 3, pp. 838-852, 2008.
[5] C. Keller, S. Hrastinski, and S. Carlsson, “Students’ Acceptance Of Pei-An Wang is a graduate student in national Chi Nan
E-Learning Environments: A comparative study in Sweden and University. He is interested in rough set theory and
Lithuani,” presented at the Proceedings of the Fifteen European e-learning platform.
Conference on Information Systems, 2007.
[6] C.-M. Chiu and E. T. G. Wang, “Understanding Web-based learning
continuance intention: The role of subjective task value,” Information
& Management., vol. 45, no. 3, pp. 194-201, 2008.
[7] M. Dougiamas and P. Taylor, Moodle: Using learning communities to
create an open source course management system, 2003.

221