AN E XAMINATION OF T HE E FFECTIVENESS OF T EACHING D ATA M ODELLING C ONCEPTS

International Journal of Database Management Systems ( IJDMS ) Vol.7, No.2, April 2015
DOI : 10.5121/ijdms.2015.7202 13
AN EXAMINATION OF THE EFFECTIVENESS OF
TEACHING DATA MODELLING CONCEPTS
Márta Czenky
Department of Informatics, Szent István University, Gödöllő, Hungary
ABSTRACT
The effective teaching of data modelling concepts is very important; it constitutes the fundament of data-
base planning methods and the handling of databases with the help of database management languages,
typically SQL. We examined three courses. The students of two courses prepared for the exam by solving
tests, while the students of the third course prepared by solving tasks from a printed exercise book. The
number of task for the second course was 2.5 times more than the number of task for the first course. The
main purpose of our examination was to determine the effectiveness of the teaching of data modelling con-
cepts, and to decide if there is a significant difference between the results of the three courses. According to
our examination, with increasing the number of test tasks and with the use of exercise book, the results
became significantly better.
KEYWORDS
Data modelling, Teaching of Data Modelling Concepts, Survey, Statistical Analysis, Data Mining Analysis
1. INTRODUCTION
At the Department of Informatics of Szent István University we teach database management for
some faculties and majors. We selected three courses where the curriculum of the subjects did not
differ, but the learning circumstances were different.
The two main topics of the Database Management subject attended by mechanical engineer stu-
dents at Faculty of Mechanical Engineering are data modelling and the SQL-92 language. The
duration of the subject is one semester with two classes per week. Classes on data modelling take
half of the semester. We supported the learning of definitions and properties of concepts and the
effective identification and realization of the concepts with tests generated from the question bank
of the Moodle e-learning education system. We included two courses in the analysis we are going
to reference these as ABK1 and ABK2. For the first course there were 92 tasks in Moodle ques-
tion bank, for the second course this number was 250. At the exam the students had to solve tests.
At Faculty of Agricultural and Environmental Sciences the subject of Computer Studies III is
taught for environment engineering students. The duration of the subject is half a semester, in-
cluding two classes of lectures and two classes of seminars per week. In the first half of the se-
mester the students learn data modelling at the lectures and SQL-92 language at the seminars. In
the second part of the semester they learn CAD knowledge. A printed data modelling exercise
book helped their preparation, but they had to solve a test at the exam [1]. Beside the exercise
book, a practice test was also available for students, consisting of 25 randomly selected tasks, so
they can try this kind of exam. We refer to this course as KM3 in the article.
We summarized in Table 1 the characteristics of the subjects. All students of the courses partici-
pated in the analysis therefore we can consider the results representative.

14
Table 1. The characteristics of subjects
Name of
subject
Abbre-
viation
Major
Level/
Course
Class per
week
lecture+
practise
Head-
count
Database
Management
ABK1
ABK2
mechanical
engineer
BSc, full
time
0+2
64
19
Computer
Studies III.
KM3
environmental
engineer
BSc, full
time
2+2 71
We are going to compare the results of electronic tests for the three courses. Based on this, we are
going to decide whether our teaching methods are efficient enough, and if there is a significant
difference between the results of the courses.
Our questions are the following:
• Which areas of concept learning can be considered good enough? We regard those areas
sufficient where the students achieved satisfactory or better result.
• In which topics and task types are the results below average?
• Did the increase in the numbers of tasks and tests improve the results significantly?
• Is there a significant difference between the effectiveness of different learning methods,
such as computational tests or printed exercise book?
• Can we determine the expected score of the student by task type and topic?
2. TEACHING OF CONCEPTS
The concept is a mental form developed from the main characteristics of the things in mind [2],
[3]. In our course at the constitution of the concept we disregard the irrelevant features and we
unite the essential characteristics into a standardized form. The content of the concept is the total-
ity of the substantive characteristics. The extent of the concept is given by the things, which dis-
pose the substantive characteristics creating the concept.
The main topics of data modelling are: basic data modelling concepts, the entity relationship
model, the relational model, dependencies, dependence diagrams and normalization. Based on
these we can classify the concepts of the data modelling into the following groups:
• The basic concepts of data modelling are: entity and set of entities, relationship and occur-
rence of relationship, attribute and value of attribute. These concepts also appear at data
models, but they are not listed separately there.
• Entity relationship model (E-R): hierarchy of set of entities, ISA relationship, inheritance
of attributes, integrity constraints.
• Relational model: relational table, keys, special data values, integrity constraints and their
enforcements, objects of relational database (index, view, synonym), relational operations.
• Dependences: functional and multivalued dependence.
• Dependence diagrams: the depiction of relational dependences of a table on diagram.
• Normalization: normal forms [4], [5], [6].
The concepts may be objective concepts, for example entity, set of entities, relational table, pri-
mary key, etc., relational concepts, for example relationship, occurrence of relationship, etc. and
operational concepts, for example relational operations, cascade update and deletion, etc.

15
The order of the teaching of topics may differ; implicit teaching of them begins with the teaching
of the basic data modelling concepts. Afterwards you may teach entity relationship model or rela-
tional model, finally dependences and normalization. [7] follows the first way while [8] the sec-
ond one. Currently we teach the relational model first, following this the entity relationship model
and its transcription into relational model. Based on our experience the entity relationship model
is more understandable for students than relational model. If we begin teaching with E-R model
though, then at one point we have to stop with it, and we can only introduce the transcription of
the E-R model into relational model after teaching relational model.
There are more ways to introduce the concepts:
• the inductive way, that is to get from the concrete to the abstract,
• the deductive way, that is to get from abstract to the concrete,
• the constructive way, to produce a representative, then the generalization of the procedure.
While teaching the concepts we primarily follow the deductive way, but the discussion of rela-
tional model happens inductively. In the database system we introduce the tables filled up with
data, their structure and relationships, after this we define the essential characteristics of the con-
cepts of relational model. [9] describes constructive way of teaching of the relational model.
The efficient concept learning means the acquirement of the following knowledge and abilities:
• the knowledge of the definition of the concept,
• concept identification, that is, you can decide whether a thing belongs to a certain concept
or not,
• concept realization, that is the recognition and enumeration examples,
• knowledge of the attributes of the concept,
• concept systematization, that is the knowledge of the relationship between concepts,
• concept classification, that is placement the concept into the hierarchical system of the con-
cepts; the concept hierarchy shows the subordinate, the superior or the coordinate relation-
ship with each other,
• the usage of the concept for the description of situations, and for the solution of problems
[10], [11], [12], [13].
[14] emphasizes to the importance of the concept learning in database planning. He emphasizes
the importance of correct concept identification and the knowledge of the attributes of the con-
cepts, which are necessary for the categorization of basic concepts. [15] also emphasizes the im-
portance of these two steps of concept learning in connection with the teaching of the E-R model.
The tasks of the Moodle question bank and exercise book used by us support the listed steps of
concept learning. The main part of tasks is smaller modelling task with which the students can
practise problem solving.
We designed separate sheets for the execution of the concept systematization and classification.
At first, students had to fill in Venn diagrams and hierarchical figures of them but lately we
changed it to concept maps of data modelling, drawn by us, which reflect the relationships be-
tween concepts more complex than the concept hierarchies [13].
3. THE STUDENTS' JUDGEMENT OF CONCEPT LEARNING
A survey with 62 questions was made among the students of ABK1 course at the end of the se-
mester to get to know the opinion of students about education of database management, including
their view about concept learning. We designed the questionnaire by instruction of [16], 54 stu-

16
dents filled it. It was random who did not fill in the survey so the survey can be considered repre-
sentative.
In the next three tables we present the answers given for the questions related to data modelling.
In the questionnaire we asked the students what kind of activities meant trouble for them in the
course of the class work? The answers are summarized in Table 2. According to the students'
opinion the most problematic activity is the comprehension of algorithms, followed by the under-
standing of concepts and definitions. It causes fewer problems to find the examples, but this con-
tradicts our teacher experiences.
Table 2. Problematic activities during class work
Activity
Proportion of
students indicat-
ing problems
usage of terminology 5.54%
understanding of the concepts and the definitions 20.36%
understanding of the example 12.99%
finding of the example 7.41%
comprehension of the algorithm 22.21%
application of the algorithm 16.63%
We indicated in Table 3 the percentage of students having problems in understanding different
concepts of data modelling .The most problematic concepts are normalization, the normal forms
and the different dependences. From the basic data modelling concepts relationship and occur-
rence of relationship are mentioned as problematic, from the relational model concepts the for-
eign key and the integrity constraints seemed to cause a problem.
Table 3. Proportion of students indicating problems with the comprehension of concepts
Concept
Proportion of
students indicat-
ing problems
entity, set of entities 5.51%
occurrence of relationship, relationship 20.38%
attribute, value of attribute 5.54%
primary key 7.39%
foreign key 14.84%
Null value 1.85%
functional dependence 44.44%
multivalued dependence 42.63%
normal forms 68.55%
integrity constraints 22.25%
The activities listed in the Table 4 belong to the area of concept identification and concept reali-
zation. Students were asked how problematic they had found the recognition of the most impor-
tant concepts of the data modelling: that is, the recognition of the basic modelling concepts and
the different dependences. The answers show that recognition of the dependences was much more
problematic than recognition of basic modelling concepts. Although students do not indicate ma-
jor problems with recognition of relationships, our teacher experience is that this is much more
complicated for them, than the recognition of the sets of entities and attributes [17].

17
Table 4. The proportion of students indicating problems with the recognition of the concepts
Activity
Proportion of
students indi-
cating problems
recognition of the sets of the entities 5.51%
recognition of the relationships 9.26%
recognition of the attributes 7.39%
recognition of the functional dependences 29.66%
recognition of the multivalued dependences 24.10%
We examined too if there is any difference between the answers of the students who achieved
better or worse result. We found that the students, who achieved better results, considered the
more easily understandable concepts less problematic but found difficult concepts problematic in
a much higher proportion than those students who received worse mark. The explanation to this
could be that students achieved better marks understood and learned the concepts better, and at
the end of the semester they were more likely to know which concept the respective questions
applied to.
4. E-LEARNING AND EXERCISE BOOK IN THE CONCEPT TEACHING
The learning of the concepts was supported by tests created from tasks of the Moodle course
management system (ABK courses), and the tasks of the printed data modelling exercise book
(KM3 course). The measurement of concept comprehension was done through electronic exam
tests. The exam tests of ABK1 and KM3 courses were randomly generated from the question
bank so every student solved different test. The exam tests of ABK2 course consisted of the same
questions for all, so every student solved the same tasks. Although the exam tests were diverse,
their tasks were from the same topic and they included the same type of task.
The practising and exam tests of the ABK1 course were different from the similar test of ABK2
course. For the ABK2 course we increased the number of tasks in the question bank, as well as
the number of the practicing and exam tests. The students of the KM3 course prepared for the
exam with the help of a printed data modelling exercise book. The exercise book includes the
tasks of the question bank of ABK2 course, and also included some new tasks. The exam test of
KM3 course was different from the exam tests of the first two courses. We made a practise test
for KM3 course so that the students could try this exam method.
Table 5. The number of tasks of question bank/exercise book and tests
Course
Preparation
method
Number of tasks in question
bank/exercise book
Number
of prac-
tice tests
Number
of exam
tests
ABK1 test 92 5 1
ABK2 test 250 14 6
KM3 exercise book 300 1 1
We summarized in Table 5 the preparation methods of the groups, the number of the tasks in the
question bank/exercise book and the number of practising and exam test.
The tasks of the tests and the exercise book require the knowledge of definitions, concept identi-
fication and knowledge of concept characteristics. At concept realization, the listing examples can
only be evaluated manually by the teacher. Therefore we always asked the enumeration of the
examples at the class. In the tests we coupled those tasks to the concept realization where exam-
ples for a concept had to be recognized.

18
Table 6. Abbreviations used in figures
Topic
Abbre-
viation
Task
Abbre-
viation
Basic modelling con-
cepts BAS giving definitions DEF
E-R model ERM knowledge of the concept characteristics CHAR
Relational model REM concept identification IDEN
Dependences DEP concept realization REAL
Normalization NOR concept collation COL
Dependence diagrams DEPD mapping of different model's concepts MAPM
mapping of concepts of different depiction MAPD
Furthermore, there are also tasks for the measurement of concept learning:
• In the course of learning basic modelling concepts, concept collation tasks includes tasks
like identification of relationships between sets of entities and mapping of attributes to a
set of entities or to a relationship. In the course of the learning of relational model con-
cepts, collation tasks were to answer what kind of basic modelling concept (set of entity,
relationship, both) is shown in the table.
• In E-R modelling at the mapping of different concepts of models the task was transcrip-
tion of E-R diagrams into relational models.
• In the case of dependence diagrams at the mapping of the concepts of the different depic-
tion the task was the mapping of different dependences described with dependence dia-
gram and schema.
We classified all tasks in the Moodle question bank by topic and by type of concept learning,
which allowed us to summarise and examine the results by topic and task type. In certain figures
in this article we used abbreviations to spare space. These abbreviations are summarized in Table
6.
Figure 1. Distribution of the tasks by topics
Figures 1 and 2 show the distribution of the tasks of the electronic tests by topic and concept
learning task type, respectively. Figure 1 shows that the four emphasized topics are: basic model-
ling concepts, relational model, dependences and normalization. These topics are the ones ensur-
ing the necessary comprehension for the planning and management of relational databases. The
E-R model and the dependence diagrams provide an alternative in database design.

19
Figure 2. Distribution of the tasks by task types
The distribution of tasks of the exercise book by topic and concept learning task types is similar
to the distribution of the tasks of the question bank of the ABK2 course.
According to Figure 2 most tasks are connected to concept realization. The explanation of this is
that there are seven normalization tests among the tests of the ABK2 course, with 7-8 questions
per test. Apart from tasks on the decomposition of tables, every task is a concept realization task.
Data modelling concepts have numerous attributes, whose knowledge is necessary for database
design. This explains the large number of tasks connected to the attributes of the data modelling
concepts.
We took into consideration the results of answers given to tasks of the electronic tests. We did not
ask students to evaluate the solutions of the tasks of the exercise book, so we do not have infor-
mation about it. At each task the incorrect answer is worth 0 point, the correct answer 1 point. We
did not apply penalty points, there was not negative point, but you could get fraction point for the
partly correct answer.
5. INVESTIGATIONAL METHODS
First we summarized the scores achieved in test trials by topic and task type, then dividing this
score with the number of trials, we determined the results. We show this data in a table and we
visualized the result of all tests and exam tests separately. We determined the mean and standard
deviation of data series and we highlighted the values below mean with bold and italic font style,
indicating that the concept learning is problematic in these topics and task types.
We also summarized the scores by task, and we divided the sum with the trial number of the
tasks. It was necessary due to the use of random questions, which means that different questions
were included in the tests variable times. Afterwards we determined the means and standard de-
viations at every data series than we examined with Chi-square test if the data have normal distri-
bution.
Between data series totalized by task we made homogeneity test by pairs to decide if the results
have same distribution, or there is a significant difference between them. If the data series have
normal distribution, than knowing that the data series are independent, we verified the equality of
standard deviations with F-test. If the standard deviations were equal, then we examined the
equality of means with T-test, if not, then we used Welch-test. In case of data series with not

20
normal distribution, we examined the equality of standard deviations with Levene-test, while the
equality of means with Mann-Whitney-test. This way we could decide if the results have the same
distribution, or any of them is significantly better [18], [19], [20].
At last, from the data of the exam tests of all three courses, we tried to make decision trees with
test scores achieved by test trial and by task. The decision trees show which scores can be ex-
pected by topic and task type, which can indicate the problematic areas of concept learning [21].
6. THE RESULTS OF THE EXAMINATION AND THEIR ASSESSMENT
We summarized in Table 7 the results by topic and by task type. In some columns of the table we
represented the results of all tests (practical and exam), while in others the results of electronic
exam tests only. In the column we indicated with bold and italic font style those values which are
below the mean of values of the column.
Table 7. The results of the concept learning
Topic Task
Proportion of the good solution (%)
ABK1 ABK2 KM3
All Exam All Exam All Exam
Basic mod-
elling con-
cepts
giving definition 82.9 89 82.9
knowledge of concept char-
acteristics
63.5 71.1 64.7 66 50.3 76.5
concept identification 73.6 86.9 73.5 81 54.3 83.3
concept realization 52.1 65.6 56.5 65.5 58.4 77.4
concept collation 69.8 77.7 78.4 64.3 61.6 84.8
E-R model
acteristics
52 65.6 76.7 63.8 56.6 76.1
mapping of the different
model's concepts
55.2 74.7 68.5 75.5 47.1 55.1
Relational
model
giving definition 79.8 89.6 77.5
acteristics
69.6 76.4 74.4 71.5 54.5 77.5
concept realization 50.1 32
concept collation 53.6 74 47.7 52.1
Depend-
ences
acteristics
24 33.3 78.2
concept identification 55.9 65
concept realization 60.6 78.4 67.8 72.5 73.75 83.3
Normaliza-
tion
concept realization 39 57.4 70.6 62.3 46.9 79.7
Depend-
ence dia-
grams
acteristics
80.3 77.7 50.2 66.2
concept realization 83.3 58.1 89
mapping of the concepts of
the different depiction
76.4 74.3
mean of the values of the
column
60.7 72.8 70.9 67.5 55.0 75.1
Some cells of the table are empty, the reasons are the followings:

21
• We first used the Moodle system when teaching the ABK1 course. The development of the
tasks happened parallel with teaching. We did not have enough time to develop all tasks in
every topic and task type by the end of the semester.
• In exam tests of the ABK2 and the KM3 courses and in the practise tests of the KM3
course there were no tasks referring to the knowledge of the definition of the concepts. The
reason behind this is that at the tests of ABK1 course and practise tests of ABK2 course we
experienced that students learned definitions well. In the exam tests we considered the so-
lution of small modelling tasks more important.
• There was time limit for the solutions of the exam tests because the students had a SQL
exam on the same class. Therefore at the ABK2 and the KM3 courses there were no ques-
tions for every topic and task type.
By the data of Table 7 we can state the followings:
• With increasing of number of tests and questions the results were better, taking into con-
sideration the results of all tests – ABK1 all and ABK2 all data series.
• The results of the ABK2 exam tests are worse compared to the result of the ABK1 exam
test. Searching for the reason we analysed how much time did the students spend with solv-
ing the practice tests. We experienced that students of ABK2 course solved practice tests in
a shorter time. It seems that they were only eager to know the correct solutions of the tests
shown after finishing the tests. This way their modelling facility developed less, which
played a part in the result of exam tests [22].
• The result of the exam test of the students of the KM3 course who learned from exercise
book was better than the results of the exam tests of the other two courses.
• The results of all tests on the KM3 course were the worst. Students of this course solved
the only one practice test many times, but their results were extremely bad, and this influ-
ences the summarized result, too.
At which topics and task types did the students achieve bad result? Hereinafter we list those top-
ics and task types where the result is below the mean at more courses:
• basic modelling concepts: knowledge of concept characteristics and concept realization,
• E-R model: knowledge of concept characteristics and mapping of the different model's
concepts,
• relational model: concept realization and concept collation,
• dependences: knowledge of concept characteristics and concept identification,
• normalization: concept realization,
• dependence diagrams: knowledge of concept characteristics.
You can see that almost every topic and task type occur in the list. Therefore we examine further
this question with data mining methods.
We summarized the result by topic and task type and depicted them on diagrams, see Figure 3
and 4.
According to Figure 3, the result was worst in entity-relationship model and normalization topics,
while the best in basic modelling concept and dependence diagrams topics. By the opinion of the
students normalization is also the most difficult topic. The entity-relationship modelling is easily
understandable and can be taught easily based on our teaching experience. Therefore we expected
better result in the solution of these tasks.

22
Figure 3. Summarized results by topics
Figure 4. Summarized results by task type
Based on Figure 4 we can say that concept realization and mapping of the different model's con-
cepts were the most problematic activities, while students can execute giving definition and con-
cept identification the best.
To decide if the differences between results are significant, we executed homogeneity tests by the
results, summarized by tasks between the following data series:
• ABK1 all – ABK2 all,
• ABK1 exam – ABK2 exam,
• ABK1 exam – KM3 exam,
• ABK2 exam – KM3 exam.
We executed the statistical examination in R program. In all cases the applied significance level
was 0.05.
We made Chi-square test for all data series to decide if the data series have normal distribution.
Our null hypothesis is:
H0: The data series has normal distribution.

23
We reject the null hypothesis if the probability value available after the execution of the test is
smaller than significance level, keep it otherwise. Table 8 includes the results of the Chi-square
tests.
In Table 8 we also show the means and the standard deviations by data series.
Table 8. Means and standard deviations
Characteristic
Data series
ABK1
all
ABK2
all
ABK1
exam
ABK2
exam
KM3
exam
Chi-square test
p/assessment
0.8554
normal
0.00024
not normal
0.1009
normal
0.00375
not normal
0.02123
not normal
mean 0.6012 0.7002 0.7287 0.6825 0.7550
standard deviation 0.1668 0.1766 0.1692 0.1667 0.1394
According to the data of Table 8, students of the KM3 course achieved the best result in the solu-
tion of the exam test. The mean of this data series is the highest, while the standard deviation of it
is the smallest. Considering means, the result was the worst at data series including results of
solution of all tests of ABK1 course.
We decided with homogeneity tests related to means and standard deviations if the differences
between means and standard deviations are significant. Our null hypotheses are:
H0: The means are the same.
H0: The standard deviations are the same.
We reject null hypotheses if the probability results of the tests are smaller than the significance
level, otherwise we keep them. The data series do not have the same distribution if either the
mean or standard deviation, or both of them are different. The result of that data series is signifi-
cantly better, which has greater mean.
Table 9 shows the results of the homogeneity tests. According to the table, the mean of ABK1all
and ABK2 all and mean of the ABK2 and the KM3 exam tests differ significantly. Although the
standard deviations of these data series differ remarkably, we could not prove significant differ-
ence. In case of the other data series there is no significant difference between means and stan-
dard deviations.
Table 9. Result of homogeneity tests
ABK1 all
ABK2 all
ABK1 exam
ABK2 exam
ABK1 exam
KM3 exam
ABK2 exam
KM3 exam
standard devia-
tion
p/test/assessment
0.2366
Levene test
same
0.8116
Levene test
same
0.1498
Levene test
same
0.3648
Levene test
same
mean
p/test/assesment
1.516e-06
Mann-Whitney
test
not same
0.173
Mann-Whitney
test
same
0.2846
Mann-Whitney
test
same
0.02343
Mann-Whitney
test
not same
We experienced that students of the KM3 course who learned from exercise book and achieved
better result learned in groups and they also made notes in the course of consultations. This was
not typical for the students who solved practise tests. Students of this course were more motivated
then the students of the ABK courses. We asked them which method they prefer. By their an-
swers the proportion is near same who prefer one or the other preparation manner [23].

24
We made the data mining examinations with open source AlphaMiner program. With the scores
achieved by task and test trials of the ABK1 all data series, we made a decision tree by topic and
by task type, to determine what scores are expected. See Table 10 and 11.
Table 10. Decision tree by task types, ABK1 all
Task Score
Confidence
(%)
Support
(%)
concept collation 1 28.7 4.8
giving definitions 1 68.7 19.1
knowledge of concept characteristics 1 44.6 31.1
concept identification 1 47.8 9.3
concept realization 0 40.6 27.1
mapping of the different model's concepts 1 33.6 8.6
Table 10 shows that the least efficient step of concept learning is concept realization. In case of
other task type the expected score is 1 point.
Table 11. Decision tree by topics, ABK1 all
Topic Score
Confidence
(%)
Support (%)
E-R model 1 32.0 17.8
Dependences 1 23.7 4.7
Basic data modelling concept 1 47.0 28.2
Normalization 0 51.1 26.8
Relational model 1 64.9 22.5
According to Table 11 you can expect the worst result at normalization, in the other topics the
probable score is 1 point. For the ABK2 course the program could not make similar decision tree.
It means that at every task type and topic the expected scores are the same. Because in the tests
solutions most of the scores are 1 point (see frequency diagram of the Figure 5), probably the
expectable score is 1 point in every task type and topic.
From the exam result of the KM3 course, we could make a decision tree by the number of tasks.
This enables us to determine the expected scores by topics because this exam test includes task in
the order of taught topics, see Table 12.
Table 12. Decision tree by number of the task, KM3 exam
Rule Score
Confidence
(%)
Support
(%)
task number: <= 14, topics: basic data modelling concepts,
relational model, E-R model 1 49.32 63.6
task number: >14 and <=16, topics: dependences, depend-
ency diagram, 0 50.83 9.1
task number: >17 and <=19, topics: normalization - primary
key determination, how many normal forms are violated by
dependences, in how many normal form is the table 0 40.53 13.6
task number: <19 ,topics: normalization - table decomposi-
tion, designation of the primary and foreign key 1 44.26 13.6
Although the values of the confidences are not high, Table 12 confirms the opinion of the stu-
dents that the most difficult topics are the dependences and the normalizations.
The results from the decision trees provided a good motivation to further analyze the scores
achieved in trial tests. What could be the reason that most of the expected scores fall into the su-

25
preme interval, and the worst expected scores fall into the bottom interval? We prepared a fre-
quency diagram by the scores of exam test of a group of the KM3 course, see Figure 5.
Figure 5. Distribution of scores achieved by questions and test trials
Figure 5 shows that 1 is the most frequent score the second most frequent is zero. The appearance
of the frequency diagram and the fact that the scores are between 0 and 1 hints that the data has
beta distribution. With Kolmogorov-Smirnov test we made a matching examination for homoge-
neity tests [24].
In the analyses undertaken it was observed that at zero value the difference is big, which is caused
by the characteristics of the beta distribution function (at zero its value is also zero), but the calcu-
lated values differ from zero because there are a lot of results with zero points. We substituted the
zero values with a small non-zero value of 0.05. It can be considered zero and it overcomes the
experienced big difference. So we succeed to prove that scores by question of homogeneity test
have generally beta distribution [25].
7. CONCLUSIONS
How effective is the concept teaching? Before the examinations we stated (see introduction), that
we regarded those areas sufficient where the students achieved satisfactory or better result. Ac-
cording to the means of the exam tests, the result of all three courses reached this level. The result
of the KM3 course almost measures up to the good result, it is only lacking a few tenth. Based on
this we can say that the concept teaching is effective.
Among the topics, the teaching of dependences and normalization are most problematic. Students
achieved the best result in basic modelling concepts and dependence diagrams topics. By task
types concept realization and the mapping of the different model's concepts have the worst re-
sults, while students achieved the best result in knowledge of definitions and concept identifica-
tion.
With the help of decision trees we could determine what result can be expected in certain topics
and task types. At the problematic areas listed above, the expected score is 0, while at other areas
it is 1. The reason behind the many 1 score results is the many 1 point scores for individual tasks.
After increasing the number of tasks in the Moodle question bank and in the tests, the results of
all tests of the ABK2 course did become significantly better than the result of all test of the ABK1
course. We could not state the same for the results of the exam tests of the courses.
The mean of the result of the exam tests of those students who prepared for the exam with the
help of printed exercise book is remarkably better. Its standard deviation is lower than the similar

26
data of the results of the exam tests of students who prepared with electronic tests. The difference
of the means is significant.
In our analysis we found that the scores of homogenous tests generally have beta distribution.
This allows us to determine - based on estimated distribution - that in case of a certain test, what
results and with what probability will students achieve.
It might worth to spend further research efforts to examine the knowledge of students about the
connections between concepts, their ability to classify the concepts and to develop the hierarchi-
cal systems of the concepts. We already examined these both. In 2010, we executed a concept
systematization survey where students had to define the connection between concepts by filling in
empty hierarchical figures and Venn diagrams. Students filled in the hierarchical figures generally
well, but we cannot say the same about Venn diagrams. In the hope of better results, we devel-
oped the concept maps of data modelling. In 2013, in a concept systematization survey, we exam-
ined if the concept systematization is significantly better with the use of concept maps. Students
had to fill in the empty concept maps of the worksheet, and give the connection between con-
cepts. The result was significantly better than the result of 2010 survey [13].
Furthermore, it can also be examined how students are able to use the concepts for describing
situations and solving problems. It might be an interesting question, if the more effective concept
learning leads to increase in the effectiveness of database design (E-R modelling and normaliza-
tion). In this topic we did not carry out any examination yet. We examined the efficiency of the
normalization, primarily focusing on alternative normalization methods that can help students
normalize better than with conventional method [26].
Before the Moodle system and data modelling exercise book were available as educational tools,
the concept learning meant learning definitions, looking for some examples, perhaps filling in
tests on paper for students. The latter one can only be used once and its reproduction is difficult,
its correction requires the teacher’s contribution and the number of tasks is few. On the contrary,
the tasks of Moodle only have to be developed once, they can be solved several times and their
correction is automatic. You can also reuse the exercise book as an alternative preparation tool
many times, if students write their answers on paper. The correctness of the results can also be
checked by the students themselves based on the solution key given in the exercise book. Both
tools support the steps of the concept learning listed above. We can recommend the usage of simi-
lar educational tools and concept systematization worksheets for every teacher. Based on our
examinations, these tools make the teaching of concepts more effective.
REFERENCES
[1] Czenky, M. (2010). Adatmodellezési példatár [Exercise book of data modelling tasks], Computer-
Books, Budapest.
[2] Merriam-Webster Dictionary (2011), http://www.merriam-webster.com/dictionary.
[3] Magyar Értelmező Kéziszótár (1979). Akadémia Kiadó, Budapest.
[4] Date, C. J. (2003). An Introduction to Database Systems, 8th Edition, Addison-Wesley: New York.
[5] Czenky, M. (2005). Adatmodellezés, SQL és Access alkalmazás, SQL Server és ADO, [Data model-
ling, application of SQL and Access, SQL Server and ADO], ComputerBooks Kiadó, Budapest.
[6] Ullmann, J. D. & Widom, J. (2008). Adatbázisrendszerek – Alapvetés [Database Systems - Basics],
Panem KFT, Budapest.
[7] Murray, M. & Guimaraes, M. (2009). „Animated Courseware Support for Teaching Database De-
sign”, Issues in Informing Science and Information Technology Volume 6, 201-211.
[8] Barnes, S. & Kuzma, J. (2009). „Empirical study in teaching first-year database students”, 7th Inter-
national Workshop on Teaching, Learning and Assessment of Databases, 6 July 2009, University of
Birmingham.
[9] McNeil, D. (1990). „Teaching Database Concepts”, In: Proceedings of ICOTS 3, Dunedin, New Zeal-
and, http://iase-web.org/Conference_Proceedings.php?p=ICOTS_3_1990, 201-208.
[10] Ambrus, A. (1995). Bevezetés a matematika didaktikába [Introduction to mathematical didactics],
ELTE Eötvös Kiadó, Budapest.

27
[11] Falus, I. (ed.) (2003). Didaktika, elméleti alapok a tanulás tanításához, [Didactics, theoretical bases
for teaching of learning], Nemzeti Tankönyvkiadó, Budapest.
[12] Kátai, Z., Nyakóné Juhász, K. & Zsakó, L. (2008). „Mivel foglalkozik az Informatika szakmódszer-
tan?”, [What Informatics metholodgy deals with?], In: Pethő, A., Herdon, M. (eds.) Proceedings of In-
formatika a felsőoktatásban 2008 Konferencia, Debrecen.
[13] Czenky, M. & Kormos, J. (2014). „Concept systematization with concept maps in data modelling”,
Teaching Mathematics and Computer Science, 12/2, 149–166.
[14] Philip, G. C. (2007). „Teaching Database Modelling and Design: Areas of Confusion and Helpful
Hints”, Journal of Information Technology Education, Volume 6, 481-497.
[15] Chilton, M. A., McHaney, R. & Chae, B. (2006). „Data modelling education: The changing technolo-
gy”, Journal of Information Systems Education, 17(1), 17-20.
[16] Falus, I. & Ollé, J. (2008). Az empirikus kutatások gyakorlata [The practice of the empirical re-
searches]. Nemzeti Tankönyvkiadó, Budapest.
[17] Czenky, M. (2008). „Adatbázis-kezelés oktatás hallgatói szemmel”, [The Education of the Database
Management through Student Eyes]. In: Pethő, A., Herdon, M. (eds.) Proceedings of Informatika a
felsőoktatásban 2008 Konferencia, Debrecen. http://www.agr.unideb.hu/if2008/
kiadvany/papers/A45.pdf.
[18] Hunyadi, L., Mundroczó, Gy. & Vita, K. (1996). Statisztika, [Statistics], Aula, Budapest.
[19] Pestman, W. R. (1998). Mathematical statistics: an introduction, de Gruyter, Berlin, New York.
[20] Sheskin, D. J (2000). Handbook of Parametric and Nonparametric Statistical Procedures, New York:
Chapman & Hall/CRC.
[21] Han, J. & Kamber, M. (2006). Data mining: Concepts and Techniques. Morgan-Kaufman, New York.
[22] Czenky, M. (2010). „Moodle tesztek használatának eredményességvizsgálata”, [Efficiency analysis of
usage of Moodle tests], In: Ollé, J. (ed.) Proceedings of II. Oktatás-Informatikai Konferencia, ELTE
Eötvös Kiadó, Budapest, 59-64.
[23] Czenky, M. (2013). „Electronic Test or Printed Exercise Book? Which Individual Mode of Prepara-
tion is More Effective?”, International Journal of Engineering Pedagogy, Volume 3, Issue 1, 12-18,
http://online-journals.org/index.php/i-jep/article/view/2306, http://dx.doi.org/10.3991/ijep.v3i1.2306.
[24] Reimann, J. (1985). Valószínűségelmélet és matematikai statisztika, [Theory of probability and ma-
thematical statistisc], Tankönyvkiadó, Budapest.
[25] Czenky, M. (2014). „Moodle tesztek eredményeinek eloszlás vizsgálata”, [Analysis of distribution of
Moodle tests results], In: Ollé, J. (ed.) Proceedings of VI. Oktatás-Informatikai Konferencia, Budap-
est, 220-235.
[26] Czenky, M. (2014) „The Efficiency Examination of Teaching of Different Normalization Method”,
International Journal of Database Management Systems (IJDMS) Vol. 6. No. 2. April 2014, pp. 1-18,
http://airccse.org/journal/ijdms/papers/6214ijdms01.pdf, DOI: 10.5121/ijdms.2014.620
AUTHORS
Márta Czenky is an assistant professor in the Department of Informatics of
Faculty of Mechanical Engineering of Szent István University (Gödöllő, Hun-
gary). She is teaching informatics, programming, database management, data-
base programming in BSC and MSC qualifications at more faculties. Her re-
search area is teaching of database management, her publications deal mainly
with this topic. More books of her were published in Hungary, among others:
Data modelling, application of SQL and Access, Let’s learn informatics to-
gether, Programming in Access, Exercise book of data modelling tasks, Exer-
cise book of Excel tasks.

AN E XAMINATION OF T HE E FFECTIVENESS OF T EACHING D ATA M ODELLING C ONCEPTS

Related slideshows

More Related Content

AN E XAMINATION OF T HE E FFECTIVENESS OF T EACHING D ATA M ODELLING C ONCEPTS