-
[ w w w . m o j - e s . n e t ]
2017
Malaysian Online Journal of Educational
Sciences Volume 5, Issue 1
January 2017
Editor-in-Chief
Professor Datuk Dr. Sufean Hussin
Editors
Assoc. Prof. Datin Dr. Sharifah Norul Akmar Syed Zamri Assist.
Prof. Dr. Onur bulan
Associate Editors
Professor Dr. Omar Abdull Kareem
Associate Prof. Dr. Ibrahem Narongsakhet
Associate Prof. Dr. Mohd Yahya Mohamed Ariffin,
Associate Prof. Dr. Norani Mohd Salleh
Associate Prof. Dr. Wan Hasmah Wan Mamat
Inst. Aydn Kiper
ISSN: 2289-3024
-
Malaysian Online Journal of Educational Sciences 2017 (Volume 5
- Issue 1)
www.moj-es.net
Copyright 2013 - MALAYSIAN ONLINE JOURNAL OF EDUCATIONAL
SCIENCES All rights reserved. No part of MOJESs articles may be
reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying, recording, or
by any information storage and retrieval system, without permission
in writing from the publisher.
Contact Address:
Professor Datuk Dr. Sufean Hussin
MOJES, Editor in Chief, University of Malaya, Malaysia
-
Malaysian Online Journal of Educational Science 2017 (Volume 5 -
Issue 1)
www.moj-es.net
Message from the editor-in-chief
Malaysian Online Journal of Educational Sciences (MOJES) strives
to provide a national and international academic forum to meet the
professional interests of individuals in various educational
disciplines. It is a professional refereed journal in the
interdisciplinary fields sponsored by the Faculty of Education,
University of Malaya. This journal serves as a platform for
presenting and discussing a wide range of topics in Educational
Sciences. It is committed to providing access to quality researches
ranging from original research, theoretical articles and concept
papers in educational sciences.
In order to produce a high quality journal, extensive effort has
been put into selecting valuable researches that contributed to the
journal. I would like to take this opportunity to express my
appreciation to the editorial board, reviewers and researchers for
their valuable contributions to make this journal a reality.
Professor Datuk Dr. Sufean Hussin, University of Malaya,
Malaysia
January 2017
Editor in chief
Message from the editor
Malaysian Online Journal of Educational Sciences (MOJES) seeks
to serve as an academic platform to researchers from the vast
domains of Educational Sciences. The journal is published
electronically four times a year.
MOJES welcomes original and qualified researches on all aspects
of Educational Sciences. Topics may include, but not limited to:
pedagogy and educational sciences, adult education, education and
curriculum, educational psychology, special education, sociology of
education, Social Science Education, Art Education, Language
Education, educational management, teacher education, distance
education, interdisciplinary approaches, and scientific events.
Being the editor of this journal, it is a great pleasure to see
the success of the journal. On behalf of the editorial team of the
Malaysian Online Journal of Educational Science (MOJES), we would
like to thank to all the authors and editors for their contribution
to the development of this journal.
Assoc. Prof. Datin Dr. Sharifah Norul Akmar Syed Zamri &
Assist. Prof. Dr. Onur bulan
January 2017
Editors
-
Malaysian Online Journal of Educational Science 2017 (Volume 5 -
Issue 1)
www.moj-es.net
Editor-in-Chief
Professor Datuk Dr. Sufean Hussin, University of Malaya,
Malaysia
Editors
Associate Professor Datin Dr. Sharifah Norul Akmar Syed Zamri,
University of Malaya, Malaysia
Assist. Prof. Dr. Onur bulan, Sakarya University, Turkey
Associate Editors
Professor Dr. Omar Abdull Kareem, Sultan Idris University of
Education, Malaysia
Associate Prof. Dr. Ibrahem Narongsakhet, Prince of Songkla
University, Thailand
Associate Prof. Dr. Mohd Yahya Mohamed Ariffin, Islamic Science
University of Malaysia
Associate Prof. Dr. Norani Mohd Salleh, University of Malaya,
Malaysia
Associate Prof. Dr. Wan Hasmah Wan Mamat, University of Malaya,
Malaysia
Inst. Aydn Kiper, Sakarya University, Turkey
Advisory Board
Emeritus Professor Dr. Tian Po Oei, University of Queensland,
Australia
Professor Dr. Fatimah Hashim, University of Malaya, Malaysia
Professor Dr. Jinwoong Song, Seoul National University,
Korea
Professor Dr. H. Mohammad Ali, M.Pd, M.A., Indonesian University
of Education, Indonesia
Professor Dr. Moses Samuel, University of Malaya, Malaysia
Professor Dr. Nik Azis Nik Pa, University of Malaya,
Malaysia
Professor Dr. Richard Kiely, the University College of St. Mark
and St. John, United Kingdom
Professor Dr. Sufean Hussin, University of Malaya, Malaysia
Dr. Zawawi Bin Ismail, University of Malaya, Malaysia
Editorial Board
Emeritus Professor Dr. Rahim Md. Sail, University Putra of
Malaysia, Malaysia
Professor Dr. Abdul Rashid Mohamed, University of Science,
Malaysia
Professor Dr. Ananda Kumar Palaniappan, University of Malaya,
Malaysia
Professor Dr. Bakhtiar Shabani Varaki, Ferdowsi University of
Mashhad, Iran.
Professor Dr. H. Iskandar Wiryokusumo M.Sc, PGRI ADI Buana
University, Surabaya, Indonesia
Professor Dr. Ramlee B. Mustapha, Sultan Idris University of
Education, Malaysia
Professor Dr. Tamby Subahan Bin Mohd. Meerah, National
University of Malaysia, Malaysia
Associate Professor Datin Dr. Sharifah Norul Akmar Syed Zamri,
University of Malaya, Malaysia
Associate Professor Dato Dr. Ab Halim Bin Tamuri, National
University of Malaysia, Malaysia
Associate Professor Dr. Abdul Jalil Bin Othman, University of
Malaya, Malaysia
-
Malaysian Online Journal of Educational Science 2017 (Volume 5 -
Issue 1)
www.moj-es.net
Associate Professor Dr. Ajmain Bin Safar, University of
Technology, Malaysia
Associate Professor Dr. Habib Bin Mat Som, Sultan Idris
Education University, Malaysia
Associate Professor Dr. Hj. Izaham Shah Bin Ismail, Mara
University of Technology, Malaysia
Associate Professor Dr. Jas Laile Suzana Binti Jaafar,
University of Malaya, Malaysia
Associate Professor Dr. Juliana Othman, University of Malaya,
Malaysia
Associate Professor Dr. Loh Sau Cheong, University of Malaya,
Malaysia
Associate Professor Dr. Mariani Binti Md Nor, University of
Malaya, Malaysia
Associate Professor Dr. Mohamad Bin Bilal Ali, University of
Technology, Malaysia
Associate Professor Dr. Norazah Mohd Nordin, National University
of Malaysia, Malaysia
Associate Professor Dr.Rohaida Mohd Saat, University of Malaya,
Malaysia
Associate Professor Dr. Syed Farid Alatas, National University
of Singapore, Singapore
Dato Dr. Hussein Hj Ahmad, University of Malaya, Malaysia
Datuk Dr. Abdul Rahman Idris, University of Malaya, Malaysia
Datin Dr. Rahimah Binti Hj Ahmad, University of Malaya,
Malaysia
Dr. Abu Talib Bin Putih, University of Malaya, Malaysia
Dr. Abd Razak Bin Zakaria, University of Malaya, Malaysia
Dr. Adelina Binti Asmawi, University of Malaya, Malaysia
Dr. Ahmad Zabidi Abdul Razak, University of Malaya, Malaysia
Dr. Chew Fong Peng, University of Malaya, Malaysia
Dr. Diana Lea Baranovich, University of Malaya, Malaysia
Dr. Fatanah Binti Mohamed, University of Malaya, Malaysia
Dr. Ghazali Bin Darusalam, University of Malaya, Malaysia
Dr. Haslee Sharil Lim Bin Abdullah, University of Malaya,
Malaysia
Dr. Husaina Banu Binti Kenayathulla, University of Malaya,
Malaysia
Dr. Kazi Enamul Hoque, University of Malaya, Malaysia
Dr. Latifah Binti Ismail, University of Malaya, Malaysia
Dr. Lau Poh Li, University of Malaya, Malaysia
Dr. Leong Kwan Eu, University of Malaya, Malaysia
Dr. Madhyazhagan Ganesan, University of Malaya, Malaysia
Dr. Megat Ahmad Kamaluddin Megat Daud, University of Malaya,
Malaysia
Dr. Melati Binti Sumari, University of Malaya, Malaysia
Dr. Mohammed Sani Bin Ibrahim, University of Malaya,
Malaysia
Dr. Mohd Rashid Mohd Saad, University of Malaya, Malaysia
-
Malaysian Online Journal of Educational Science 2017 (Volume 5 -
Issue 1)
www.moj-es.net
Dr. Muhammad Azhar Bin Zailaini, University of Malaya,
Malaysia
Dr. Muhammad Faizal Bin A. Ghani, University of Malaya,
Malaysia
Dr. Nabeel Abdallah Adedalaziz, University of Malaya,
Malaysia
Dr. Norlidah Binti Alias, University of Malaya, Malaysia
Dr. Pradip Kumar Mishra, University of Malaya, Malaysia
Dr. Rafidah Binti Aga Mohd Jaladin, University of Malaya,
Malaysia
Dr. Rahmad Sukor Bin Ab Samad, University of Malaya,
Malaysia
Dr. Renuka V. Sathasivam, University of Malaya, Malaysia
Dr. Rose Amnah Bt Abd Rauf, University of Malaya, Malaysia
Dr. Selva Ranee Subramaniam, University of Malaya, Malaysia
Dr. Sit Shabeshan Rengasamy, University of Malaya, Malaysia
Dr. Shahrir Bin Jamaluddin, University of Malaya, Malaysia
Dr. Suzieleez Syrene Abdul Rahim, University of Malaya,
Malaysia
Dr. Syed Kamaruzaman Syed Ali, University of Malaya,
Malaysia
Dr. Vishalache Balakrishnan, University of Malaya, Malaysia
Dr. Wail Muin (Al-Haj Said) Ismail, University of Malaya,
Malaysia
Dr. Wong Seet Leng, University of Malaya, Malaysia
Dr. Zahari Bin Ishak, University of Malaya, Malaysia
Dr. Zahra Naimie, University of Malaya, Malaysia
Dr. Zanaton Ikhsan, National University of Malaysia,
Malaysia
Dr. Zeliha DEMIR KAYMAK, Sakarya University, Turkey
Cik Umi Kalsum Binti Mohd Salleh, University of Malaya,
Malaysia
En. Mohd Faisal Bin Mohamed, University of Malaya, Malaysia
En. Norjoharuddeen Mohd Nor, University of Malaya, Malaysia
En. Rahimi Md Saad, University of Malaya, Malaysia
Pn. Alina A. Ranee, University of Malaya, Malaysia
Pn. Azni Yati Kamaruddin, University of Malaya, Malaysia
Pn. Fatiha Senom, University of Malaya, Malaysia
Pn. Fonny Dameaty Hutagalung, University of Malaya, Malaysia
Pn. Foziah Binti Mahmood, University of Malaya, Malaysia
Pn. Hamidah Binti Sulaiman, University of Malaya, Malaysia
Pn. Huzaina Binti Abdul Halim, University of Malaya,
Malaysia
Pn. Ida Hartina Ahmed Tharbe, University of Malaya, Malaysia
-
Malaysian Online Journal of Educational Science 2017 (Volume 5 -
Issue 1)
www.moj-es.net
Pn. Norini Abas, University of Malaya, Malaysia
Pn. Roselina Johari Binti Md Khir, University of Malaya,
Malaysia
Pn. Shanina Sharatol Ahmad Shah, University of Malaya,
Malaysia
Pn. Zuwati Binti Hashim, University of Malaya, Malaysia
-
Malaysian Online Journal of Educational Science 2017 (Volume 5 -
Issue 1)
www.moj-es.net
Table of Contents
CONCEPTIONS OF THE NATURE OF SCIENCE HELD BY UNDERGRADUATE
PRE-SERVICE BIOLOGY TEACHERS IN SOUTH-WEST NIGERIA
1
Adedoyin, A. O [1], Bello, G
DIFFERENTIAL ITEM FUNCTIONING ANALYSIS OF HIGH-STAKES TEST IN
TERMS OF GENDER: A RASCH MODEL APPROACH
10
Seyed Mohammad Alavi, Soodeh Bordbar
EFFECTIVENESS OF BLENDED LEARNING AND E-LEARNING MODES OF
INSTRUCTION ON THE PERFORMANCE OF UNDERGRADUATES IN KWARA STATE,
NIGERIA
25
Amosa Isiaka GAMBARI, Ahmed Tajudeen SHITTU, O. Olufunmilola
OGUNLADE, Olourotimi Rufus OSUNLADE
THE EFFECT OF SCHOOL BUREAUCRACY ON THE RELATIONSHIP BETWEEN
PRINCIPALS LEADERSHIP PRACTICES AND TEACHER COMMITMENT IN MALAYSIA
SECONDARY SCHOOLS
37
Teoh Hong Kean, Sathiamoorthy Kannan, Prof Chua Yan Piaw
THE EFFECT OF TIME ON DIFFICULTY OF LEARNING (THE CASE OF
PROBLEM SOLVING WITH NATURAL NUMBERS)
56
Deniz KAYA, Cenk KEAN
THE RELATIONSHIP BETWEEN PROBLEMATIC INTERNET USE, ALEXITHYMIA,
DISSOCIATIVE EXPERIENCES AND SELF-ESTEEM IN UNIVERSITY STUDENTS
75
Murat Iskender, Mustafa Ko, Neslihan Arici, Naciye Gven
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Conceptions Of The Nature Of Science Held By Undergraduate
Pre-Service Biology Teachers In South-West Nigeria
Adedoyin, A. O [1], Bello, G [1]
[1] Department of Science Education, University of Ilorin,
Ilorin 240001, Nigeria *Corresponding author
[email protected], Tel. +234(0)8066762605
ABSTRACT
This study investigated the conceptions of the nature of science
held by pre-service undergraduate biology teachers in South-West,
Nigeria. Specifically, the study examined the influence of their
gender on their conceptions of the nature of science. The study was
a descriptive research of the survey method. The population for the
study comprised all undergraduate pre-service biology teachers in
Nigerian universities. Stratified random sampling technique was
used to select ninety nine (99) undergraduate pre-service biology
teachers from three universities in SouthWest, Nigeria. The nature
of science questionnaire (NoSQ) was used to collect data. Results
revealed that pre-service undergraduate teachers gender did not
influence their conceptions. It was recommended that biology
teacher educators should equip the pre-service undergraduate
biology teachers with meta-cognitive tools such as Study Technology
to enable them to learn for meaningful understanding.
Keywords: Nature of Science, Conceptions, Misconceptions,
Correct Conceptions
INTRODUCTION
Science has since the dawn of civilization been a potent tool
for finding solutions to the never ending human problems, or at
least, help man to manage his challenges well. Science as a field
of study and endeavour will always be an important aspect of human
lives. Science involves all conscious activities that man engages
in to understand nature and its components. Science, according to
Abimbola (2013), can be seen as a body of knowledge; it could also
mean a way or method of investigation and a way of thinking in an
attempt to understand nature. Amongst others, the scientific
process involves particular skills of inquiry that include:
observing, classifying, experimenting, measuring, inferring and
organizing data.
The nature of science according to GessNewsome (2002) is defined
as the epistemological foundations of science, which include its
empirical basis, tentativeness, subjectivity, creativity,
unification, and its cultural and social embedded characteristics.
The nature of science encapsulates the characteristics of science
that make people understand scientific endeavours with less
acquisition of cumbersome scientific knowledge. The preceding
descriptions of the nature of science cannot be wholesome because
science is viewed from different points of view and perspectives by
researchers and scientists the world over. If there is a list that
attempts to reveal all the values, processes, usefulness, prospects
and products of science, the list will be endless.
Scientists and science educators have emphasized the absence of
a consensus among researchers and scientists on the meaning of the
nature of science. They opined that the situation is so because
the
www.moj-es.net
1
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
nature of science is multifaceted, everchanging and convoluted.
Like scientific knowledge, conceptions of the nature of science are
ever dynamic and have witnessed different transformations
throughout the development of science and scientific processes.
Moreover, despite continuing disagreements about a particular
definition of the nature of science, at a certain level of
generality and within a set period, there is a shared perspective
about the nature of science. There is a general agreement on
several elements of the nature of science that is used for research
purposes (AbdElKhalick, 2005; AbdELKhalick and Lederman, 2008;
Akerson, Morrison & McDuffie, 2006; Lederman, 2007). The main
purpose of this study was to gain useful insight into what the
conceptions of the nature of science held by Nigerian undergraduate
preservice biology teachers are and take appropriate actions where
necessary. Specifically, the study also aimed to; find out
undergraduate preservice biology teachers conceptions of the nature
of science; determine the influence of gender on conceptions of the
nature of science held by undergraduate preservice biology teachers
in SouthWest, Nigeria.
For any science student to excel in the field of science,
adequate knowledge of the nature of science is essential. The
structure, epistemology and philosophy of science as described by
Abimbola (2013) include its products, processes, and ethics. It is
only when a student gets a good grasp of these concepts that such
student could stand a good chance of a constructive academic
success.
Literature shows that relatively little attention has been paid
to students views about the nature of science. This is more so in
the case of undergraduate preservice biology teachers (Kang,
Scharmann & Noh, 2005; elikdemir, 2006). There have been many
studies about students views of the nature of scientific knowledge,
but those conducted among the undergraduate preservice biology
teachers in Nigeria are relatively few. Meanwhile, there exist
numerous studies carried out among undergraduate preservice science
teachers and practicing science teachers in other parts of the
world (Aslan, 2009; Ayvac, 2007; Kenar, 2008; YcelOyman, 2002).
The results of these studies made it clear that the majority of
Turkish elementary school students and preservice science teachers
held misconceptions and alternative conceptions of some aspects of
the nature of science. Many of the undergraduate science education
students had the conception that there is certain and defined
scientific method to develop scientific knowledge (Blbl & Kk,
2007; ahin, Deniz & Grgen, 2006; naloban & Ergin, 2008).
However, Most of these studies have neglected the possible
influence of preservice biology teachers gender on their
conceptions of the nature of science. Also, many of these
researches were conducted involving only preservice biology
teachers from one particular university or college, whereas, this
study focused on preservice biology teachers from three different
Nigerian universities. This is the knowledge gap this study
intended to fill.
Bearing in mind the cultural diversity in the Nigerian context
and the peculiarities of the Nigerian educational system,
considerable attention should be focused at understanding what the
views of preservice biology teachers about the nature of science
are. This knowledge will help stakeholders arrive at a
comprehension of the Nigerian biology teachers conception of the
nature of science and take appropriate palliative or remedial
measures where necessary.
The theoretical foundation upon which this study was based is
constructivism. This area of educational interest is a learning
theory propounded by cognitive psychologists with constructivist
epistemological perspectives such as Jean Piaget (18961980). Other
writers and philosophers who have enormously influenced
constructivism are: John Dewey, Maria Montessori, Lev Vygotsky
among others (Wikipedia, 2014). Brooks (2004) defined the term
constructivism as an instructional approach that emphasises the
active participation of learners in the instructional process.
Effective learning takes place, as learners are active components
of a process of meaning and knowledge construction instead of
passively receiving information. Brooks (2004) perceived
constructivism as a learning theory based on observation and
scientific study. The researcher also went further to explain it as
a philosophy of learning founded on the belief that by reflecting
on human experiences, man can build up an understanding of the
world. Learning could also be perceived as the process of adjusting
our mental states to accommodate new experience (Brooks, 2004).
Read (2004) opined that, there is widespread agreement among
researchers in education that learners should not be seen as
passive recipients of information during the teaching and learning
process. Instead, they are active constructors of their knowledge.
He explained that before a child begins school, he has a wealth of
experiences, and these prior experiences have led him to develop a
common sense
www.moj-es.net
2
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
understanding of his social and natural environment. These
experiences give the learning process a boost because the formation
of new knowledge will likely be influenced by preexisting
knowledge. However, there may be a challenge that could arise from
learners personal construction of new knowledge.
The study was guided by the following research questions; 1.
What are undergraduate preservice biology teachers conceptions of
the Nature of Science? 2. Is there difference in the number of
correct conceptions about the nature of science held by male
and female undergraduate preservice biology teachers? 3. Is
there difference in the number of misconceptions about the nature
of science held by male and
female undergraduate preservice biology teachers? Based on the
preceding research questions, it was hypothesised that; HO1: There
is no significant difference in the number of correct conceptions
about the nature of
science held by male and female undergraduate preservice biology
teachers. HO2: There is no significant difference in the number of
misconceptions about the nature of
science held by male and female undergraduate preservice biology
teachers. The outcome of this study is envisaged to be of
importance to the teaching and learning of science.
Specifically; students, secondary school biology teachers and
lecturers in tertiary institutions, teacher educators, curriculum
planners and textbook writers.
Secondary school biology teachers and lecturers in higher
institutions of learning stand to gain immensely from the findings
of this research work. The outcome of this research might help them
realize the conceptions held by their students, and take
appropriate steps towards improving the meaningful understanding of
scientific concepts and the nature of science by the students.
On the part of biology teacher educators, the findings of this
research work could help them to tailor biology teacher education
practices towards producing biology teachers that can successfully
handle students misconceptions and improve the image of biology
among the students. Curriculum experts need to be sensitive to
students preconceptions at various stages of curriculum
development. Hence, results of this study could provide them with
useful information on students conception of the nature of
science.
Textbook writers might also find the outcome of this study
useful because it might illuminate the hidden educational needs of
secondary school science students. The upshot of this study would
keep them abreast of the conceptions of the nature of science held
by students, thereby giving them directions in which they should
make improvements in addressing students misconceptions and
alternative conceptions of the nature of science.
A review of literature related to this study was carried out.
Bello and Abimbola (1997) conducted a study to determine the
impacts of gender on students conceptmapping ability and
achievement in evolution. The upshot of the study showed that there
was no gender influence on students' conceptmapping ability and
their achievement in evolution. In Ghana, Taale (2014) also
conducted a study to inquire about the influence of biology
teachers gender on their conceptions of the nature of science. The
outcome of the study is also similar to that of Bello and Abimbola
(1997). The study also showed that teachers gender is not a major
factor in their conceptions of the nature of science.
From the information provided by the reviewed literature, it is
evident that the researchers share the opinion that there exist
various misconceptions and alternative conceptions about the nature
of science among undergraduate preservice science teachers,
especially practicing biology teachers and secondary school
students alike. Unlike what was observed in some of the reviewed
literature, where the researchers utilized simple random sampling
and in some cases, purposive sampling, the researcher in the case
of this study employed the stratified random sampling technique in
selecting the sample for the study, this is at variance with some
of the reviewed studies where the purposive sampling method was
adopted for drawing samples.
The need for this sort of research was borne out of the fact
that most researches into the conceptions of the nature of science
held by preservice biology teachers were conducted outside Nigeria
while those conducted in Nigeria are relatively few. The execution
of a research work of this nature is envisaged to help bridge the
gap created by the relatively small number researches into the
conceptions of the nature of science held by Nigerian undergraduate
preservice biology teachers. The sample that was used for the
www.moj-es.net
3
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
present study was gotten from three different universities,
unlike other past researches that made use of just one institution.
This would also enhance the external validity of the study.
METHODOLOGY
The study was a descriptive research of the survey method. The
population for this study was all undergraduate preservice biology
teachers in SouthWest Nigeria. Stratified random sampling technique
was adopted to select the representation of the population. The
universities from which samples was drawn are government
universities that have a long history of graduating biology
education students. Specifically, ninety nine (99) preservice
biology teachers were selected from three universities in SouthWest
Nigeria.
The research instrument that was employed to gather data in this
study is the Nature of Science Questionnaire (NoSQ). The researcher
adapted the NoSQ from the previous instrument developed and used by
Indiana State University (2015). The NoSQ was divided into two
sections; sections A and B. Section A of the questionnaire sought
for demographic information while items in Section B sought for the
preservice biology teachers conceptions about the nature of
science. There were 25 items in Section B, some of which have been
restructured to suit this study. These items were based on the
various tenets of the nature of science. Respondents were required
to indicate their conceptions about the nature of science by
ticking () or crossing (X) the various statements about the tenets
of the nature of science. A reliability coefficient of 0.74 was
obtained using Pearson product moment correlation statistics.
Both descriptive and inferential statistics were employed in the
analysis of the collected data. All the research questions raised
and hypotheses earlier stated were tested using the chisquare (2)
statistical tool. The hypotheses were tested at 0.05 level of
significance.
RESULTS
Research Question 1: What are undergraduate pre-service biology
teachers conceptions of the Nature of Science?
To answer research question 1, undergraduate preservice biology
teachers were requested to indicate their conceptions of the nature
of science. Table 1 and table 2 show the number, as well as
percentages of undergraduate preservice biology teachers that held
correct conceptions and misconceptions about various aspects of the
nature of science respectively. As shown in table 1 and table 2,
undergraduate preservice biology teachers held a mixture of correct
and misconceptions about the nature of science. However, they
appeared to hold more misconceptions than correct conceptions about
the nature of science. This finding provides the answer to research
question 1.
Table 1. Correct Conceptions about the Nature of Science held by
Undergraduate Pre-Service Biology Teachers.
S/N Pre-service undergraduate biology teachers misconceptions
about the nature of science Correct Conceptions
Frequency Percentage
1 Science is primarily concerned with understanding how the
natural world works. 78 79.4%
2 Science requires a lot of creative activity. 77 78.6% 3
Science typically provides only temporary answers to questions. 50
51.0% 4 Scientists can believe in God or a supernatural being and
still do good science. 64 64.6%
5 Science can be done poorly. 40 40.4% 6 Science can study and
explain events that happened millions of years ago. 72 73.5%
7 Knowledge of what science is, what it can and cannot do, and
how it works, is important for all educated people. 85 85.9%
8 Scientists have observed that nature apparently follows the
same rules throughout the universe. 65 65.7%
www.moj-es.net
4
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
9 Scientists often try to test or disprove possible
explanations. 78 78.8%
10 Science can be influenced by the race, gender, nationality,
or religion of the scientists. 55 55.6%
Table 2. Misconceptions about the Nature of Science held by
Undergraduate Pre-Service Biology Teachers.
S/N
Pre-service undergraduate biology teachers misconceptions about
the nature of science
Misconceptions Frequency Percentage
1 Science is primarily a search for truth. 85 86.7% 2 Science
can solve any problem or answer any question. 66 66.7% 3 Science
can use supernatural explanations if necessary. 36 36.4%
4 Astrology (predicting your future from the arrangement of
stars and planets) is a science. 66 66.7%
5 A hypothesis is an educated guess about anything. 73 74.5% 6
Science is most concerned with collecting facts. 79 82.3% 7 Most
engineers and medical doctors are actually scientists. 83 83.8%
8 A scientific fact is absolute, fixed, and permanent. 59
59.6%
9 A scientific theory is a guess. 47 48.5% 10 Scientists have
solved most of the major mysteries of nature. 67 67.7%
11 Modern scientific experiments usually involve trying
something to see what will happen, without predicting a likely
result. 71 74.7%
12 Anything done scientifically is always accurate and reliable.
69 69.7% 13 All scientific problems must be studied with The
Scientific Method. 73 73.7% 14 Disagreement between scientists is
one of the weaknesses of science. 45 45.9% 15 Any study done
carefully and based on observation is scientific. 73 73.7%
Research Question 2: Is there difference in the number of
correct conceptions about the nature of
science held by male and female undergraduate pre-service
biology teachers? HO1 ; There is no significant difference in the
number of correct conceptions about the nature of
science held by male and female undergraduate pre-service
biology teachers.
Table 3: Chi-square Analysis of Significant Difference in the
Number of Correct Conceptions held by Male and Female Pre-service
Undergraduate Biology teachers
Not Significant at .05 alpha level of significance. As shown in
Table 3, a chisquare analysis was conducted to compare the correct
conceptions about
the nature of science held by male and female undergraduate
preservice biology teachers. It was found that there was no
significant difference in the number of correct conceptions about
the nature of science held by male and female undergraduate
preservice biology teachers. [(1, 99) = 25.296, p =.235]. Since the
pvalue (.235) is greater than 0.05 (level of significance), the
null hypothesis (HO1) was not rejected. This finding provides an
answer to research question 1. That is to say, there is no gender
difference in the number of correct conceptions about the nature of
science held by undergraduate preservice biology teachers. This
Gender Df Sig
Pearson ChiSquare 25.296 1 .235
Likelihood Ratio 32.165 1 .056 LinearbyLinear Association .326 1
.568
N of Valid Cases 99
www.moj-es.net
5
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
result suggests that gender really does not have influence on
the number of correct conceptions of the nature of science held by
undergraduate preservice biology teachers.
Research Question 3: Is there difference in the number of
misconceptions about the nature of science
held by male and female undergraduate pre-service biology
teachers? HO2 ; There is no significant difference in the number of
misconceptions about the nature of science
held by male and female undergraduate pre-service biology
teachers.
Table 4:Chi-square Analysis of Significant Difference in the
Number of Misconceptions held by Male and Female Pre-service
Undergraduate Biology teachers
Not Significant at .05 alpha level of significance. A chisquare
analysis was conducted to compare the number of misconceptions
about the nature of
science held by male and female undergraduate preservice biology
teachers. As shown In Table 4, There was no significant difference
in the number of misconceptions about the nature of science held by
undergraduate preservice biology teachers based on their gender at
the p>. 05 level [(1, 99) = .009, p =.923]. Since the pvalue
(.923) is greater than 0.05 (level of significance), the null
hypothesis (HO2) was not rejected. This finding answers research
question 2, meaning there is no difference in the number of
misconceptions about the nature of science held by undergraduate
preservice biology teachers based on their gender. This result
indicates that undergraduate preservice biology teachers gender
does not have much influence on their conceptions of the nature of
science.
DISCUSSION
Findings from the study showed that undergraduate preservice
biology teachers held both correct conceptions and misconceptions
about the nature of science. The finding suggests that
undergraduate preservice biology teachers held more misconceptions
about the nature of science than correct conceptions. This is in
line with the works of Butler et al. (2014), Hanson (2015),
Onijamowo (2010), Sangsaard et al. (2014), Stojanovska,
Soptrajanov, and Petrusevski (2012), Tan and Taber (2009),
Pinarbasi, Sozbilir, and Canpolat (2009) all these researchers
claimed that there exist numerous misconceptions among practicing
and preservice biology teachers about the nature of science.
The gender difference in the conception of undergraduate
preservice biology teachers was found to be statistically
insignificant. This outcome agrees with those of Oluwatayo (2011),
Taale (2014) who concluded that there is no significant difference
in the number of biology teachers who held correct conceptions and
misconceptions about the nature of science regarding gender. Parts
of the reasons adduced for this is that both male and female
undergraduate preservice biology teachers are trained by the same
teachers; under the same conditions; and are taught using the same
curriculum. Hence, there is little or no room for a variance in
their conceptions.
Consequent upon the fact that there exist various misconceptions
about the nature of science among preservice biology teachers which
spreads across both genders as reported in this study, the teaching
of science and biology in particular in secondary schools and
institutions of higher learning is at a disadvantage. Olorundare
(2014b) reported that students usually experience difficulty in
learning science topics because of the misconceptions held by their
science teachers which are easily transferred to the students.
Misconceptions are understood to be stubborn and resistant, hence
the higher risk of learners, carrying along the same misconceptions
about the nature of science through their elementary education till
graduation.
Gender Df Sig
Pearson ChiSquare .009 1 .923 Likelihood Ratio .009 1 .923
LinearbyLinear Association .009 1 .924
N of Valid Cases 99
www.moj-es.net
6
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
The implication of this is that such graduates eventually become
teachers and transfer same
misconceptions about the nature of science on to another
generation of learners which will make the task of overcoming
students failure in science subjects and biology in particular
difficult if not impossible. The findings of this study should
stimulate the education authorities to proactively device methods
to arrest the unwanted level of misconceptions among preservice
biology teachers and enable them to effectively educate their
students for meaningful understanding of biology and science
generally.
CONCLUSION
The study concluded that undergraduate preservice biology
teachers held both correct conceptions and misconceptions about the
nature of science; however, they held more misconceptions than
correct conceptions about the nature of science. The study further
concluded that the gender of preservice biology teachers did not
influence their conceptions about the nature of science. This study
has shed more light on the conceptions of the nature of science
held by undergraduate biology teachers in Nigeria. Irrespective of
preservice biology teachers gender, they exhibited similar
conceptions of the nature of science. The high number of
misconceptions held by these teachers implies that remedial
measures that will help these biology teachers reconcile their
misconceptions about the nature of science with the appropriate
scientific conceptions of the nature of science should be
prioritized. Also, the study has pointed out the urgent need for
capacity building programs on the nature of science for biology
teachers in both government and private schools. These programs may
include; symposia, seminar and workshops. The findings of this
study conform to the international realities of the existence of
various misconceptions and alternative conceptions of the nature of
science among biology teachers in many countries. This being the
case, international agencies with focus on the advancement of
science and biology in particular should focus more attention on
the minimization and possible eradication of misconceptions among
biology teachers in Nigeria. This could be through the introduction
of international scientific literacy exchange programs for both
preservice and inservice biology teachers.
RECOMMENDATIONS
The following recommendations are considered relevant based on
the findings of this study: 1. Biology teacher education curriculum
planners should take cognisance of the fact that there exist
numerous misconceptions about the nature of science among
preservice biology teachers; hence, there is a need to introduce
the nature of science as a separate course in the Nigerian biology
teacher education curricula.
2. There is also a need to retrain practicing biology teachers
to help them reconcile their misconceptions about the nature of
science with the appropriate scientific conceptions of the nature
of science. This will prevent such teachers from passing
misconceptions to their students in the science classroom.
3. Biology teacher educators should regularly identify the
preservice undergraduate science teachers' misconceptions about the
nature of science and take appropriate pedagogical measures to help
them to reconcile the misconceptions with the appropriate
scientific conceptions
4. Biology teacher educators should equip the preservice
undergraduate science teachers with metacognitive tools such as
Study Technology to enable them to learn how to learn for
meaningful learning.
5. Biology education programmes should include Misconceptions in
science a core course. This will help the biology teachers and
students to be more sensitive to misconceptions about scientific
concepts and how to avoid and reconcile misconceptions with the
appropriate scientific conceptions.
6. Biology textbooks writers should also take note of
misconceptions about the nature of science, hence, guide against
statements and assertions that encourage misconceptions.
7. Biology teacher training programmes should make room for the
use of instruments such as the nature of science questionnaire as
formative tools. This would improve undergraduate biology
www.moj-es.net
7
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
teachers awareness of the nature of science and aid their
understanding of the processes of scientific inquiry and the
scientific enterprise. This study was specifically carried out on
the conceptions of the nature of science held by preservice
biology teachers in SouthWest Nigeria, this kind of study can be
carried out in other parts of the country to give a holistic view
of what the conceptions of the nature of science held by Nigerian
undergraduate preservice biology teachers are. Variables not
covered in this study can also be investigated by other
researchers. Further studies can also be conducted to look into the
sources of the misconceptions or correct conceptions of the nature
of science held by biology teachers in Nigeria. Researches can also
be conducted to compare the conceptions of the nature of science
held by preservice and inservice biology teachers to see if their
conceptions of the nature of science changes with their teaching
experience.
More researches can also be carried out to determine if there is
any relationship between preservice biology teachers conceptions of
the nature of science and their academic achievement in biology
courses. Also, this can be replicated among inservice biology
teachers to find out if their conceptions of the nature of science
have anything to do with their classroom instruction or the
academic achievement of their students in biology.
REFERENCES
AbdElKhalick, F. (2005). Developing deeper understandings of
nature of science: The impact of a philosophy of science course on
preservice science teachers' views and instructional planning.
International Journal of Science Education, 27(1), 1542.
AbdElKhalick, F., & Laderman, N. G. (2008). Improving
science teachers conceptions of the nature of science: A critical
review of literature. International Journal of Science Education,
22(7), 665701.
Abimbola, I. O. (2013). Philosophy of science for degree
students. Ilorin: Bamitex printing & publishing.
Akerson, V., Morrison, J., & McDuffie, A. (2006). One course
is not enough: Pre service elementary teachers; retention of
improved views of nature of science. Journal of Research in Science
Teaching, 43(2), 194213.
Aslan, O. (2009). Science and technology teachers views on
nature of science and the reflexions of these views on classroom
activities. Published doctoral dissertation, Gazi University,
Ankara.
Ayvac, H. . (2007). A study toward teaching the nature of
science based on different approaches for classroom teachers in
gravity content. Published doctoral dissertation, Karadeniz
Technical University.
Brooks, J. G. (2004). Workshop: Constructivism as a paradigm for
teaching and learning. Educational Broadcasting Corporation.
Retrieved 16/9/2014 from
http://www.thirteen:org/edonline/concepttoclass/constructivism
ndex
Blbl, K. & Kk, M. (2007). Investigating elementary students
view about scientific knowledge. Primary School Congress: Primary
school education bulletin booklet, Hacettepe University,
Ankara.
Butler, J., Simmie, G. M. & OGrady, A. (2014). An
investigation into the prevalence of ecological misconceptions in
upper secondary students and implications for preservice teacher
education. European Journal of Teacher Education. 14(2)22 Retrieved
from: www.emeraldinsiight.com. doi:
10.1080/02619768.2014.943394.
elikdemir, M. (2006). Examining middle school students
understanding of the nature of science. Published master thesis,
Middle East Technical University, Ankara.
GessNewsome, J. (2002). The use and impact of explicit
instruction about the nature of science and science inquiry in an
elementary science methods course. Science and Education, 11(15),
5567.
www.moj-es.net
8
http://www.thirteen:org/edonline/concepttoclass/constructivism%09ndex
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Hanson, R. (2015). Identifying students alternative concepts in
basic chemical bonding: A case study of
teacher trainees in the University of Education, Winneba.
International Journal of Innovative Research and Development, 4 (1)
115-122.
Indiana State University (2015). Science knowledge survey.
Retrieved from www.indiana.edu/../sci.tst.html
Kang, S., Scharmann, L. C., & Noh, A. (2005). Examining
students views on the nature of science: Results from Korean 6th,
8th and 10th graders. Science Education, 89(13), 314334.
Kenar, Z. (2008). Prospective science teachers views of the
nature of science. Published master thesis. Balkesir University,
Balkesir. Khata.
Lederman, N. (2007). Nature of science: Past, present, and
future. In S. L. Abell, N. (Ed.), Handbook of Research on Science
Education. (3, pp.18). Mahwah: Lawrence Erlbaum Associates.
Olorundare, A. S. (2014b). Theory into practice: beyond surface
curriculum in science education. Proceeding of 147th Inaugural
lecture, University of Ilorin.
Oluwatayo J. A. (2011). Gender difference and performance of
secondary school students in mathematics. European Journal of
Educational Studies 3(1), 9495.
Onijamowo, O. T. (2010). Senior School Chemistry student
misconceptions and alternative conceptions of selected chemistry
concepts in Kogi State Nigeria. Unpublished master Dissertation.
University of Ilorin, Nigeria.
Pinarbasi, T., Sozbilir, M. & Canpolat, N. (2009).
Prospective chemistry teachers misconceptions about colligative
properties: Boiling point elevation and freezing point depression,
Chemical Education Research and Practice, 9(10), 273280.
Read, J. R. (2004). Childrens misconceptions and conceptual
change in science education. Retreived from:
http://acell.chem.usyd.edu.au/ConceptualChange.cfm.
ahin, N., Deniz, S., & Grgen, I. (2006). Student teachers
attitudes concerning understanding the nature of science in turkey.
International Education Journal, 7 (1), 5155.
Sangsaarda, R., Thathongb, K. & Chapoo, S. (2014). Examining
grade 9 students conceptions of the nature of science. Procedia
Social and Behavioral Sciences, 116(42), 382 388. Retrieved from
www.sciencedirect.com.
Stojanovska, M., Soptrajanov, B. & Petrusevski, V. (2012).
Addressing misconceptions about the particulate nature of matter
among secondary school and highschool students in the Republic of
Macedonia. Creative Education, 3, 619631.
doi:10.4236/ce.2012.35091.
Taale, K. D. (2014). Gender and location influence on Ghanaian
students perceptions of energy and classroom learning.
International Journal of Education and Practice, 2(3), 5166.
Tan, K. D. & Taber, K. (2009). Ionization energy:
Implications of preservice teachers' conceptions. Journal of
Chemical Education, 86(5), 623629.
Wikipedia (2014). constructivism. Retrieved September 22 2014 at
6:19pm. From http://wikipedia.com/constructivism.
www.moj-es.net
9
http://www.indiana.edu/sci.tst.htmlhttp://acell.chem.usyd.edu.au/ConceptualChange.cfmhttp://www.sciencedirect.com/http://dx.doi.org/10.4236/ce.2012.35091http://wikipedia.com/constructivism
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Differential Item Functioning Analysis of High-Stakes Test in
Terms of Gender: A Rasch Model Approach
Seyed Mohammad Alavi [1], Soodeh Bordbar [2]
[1] University of Tehran, Tehran, Iran
Email:[email protected] [2] PhD candidate of TEFL, Tehran
University, Tehran, Iran Email: [email protected]
ABSTRACT
Differential Item Functioning (DIF) analysis is a key element in
evaluating educational test fairness and validity. One of the
frequently cited sources of construct-irrelevant variance is gender
which has an important role in the university entrance exam;
therefore, it causes bias and consequently undermines test
validity. The present study aims at investigating the presence of
DIF in terms of gender in a high stakes language proficiency test
in Iran, the National University Entrance Exam for Foreign
Languages (NUEEFL). The participants responses (N = 5000) were
selected randomly from a pool of examinees who had taken the NUEEFL
in 2015. The results displayed DIF between male and female test
takers. Hence, on the basis of the findings, it is concluded that
the NUEEFL test scores are not free of construct-irrelevant
variance and the overall fairness of the test is not confirmed.
Also, both Rasch assumptions (i.e., unidimensionality and local
independence) are hold in the present research.
Keywords: Differential Item Functioning, Dimensionality, Rasch
Model
INTRODUCTION
In language testing and educational measurement the discussions
about test use and the consequences of tests have increased. Since
the National University Entrance Exam for Foreign Languages
(NUEEFL) is administered annually to a large number of test takers
country-wide in Iran, the consequences of failure on the test are
serious. It could result in spending one or more years for test
preparation and two-year military service (for males).
Therefore, it is essential to examine the extent to which the
instrument assesses what it is intended to measure (validity) as
well as the test consistency (reliability) (Pae, 2011) in measuring
the English ability in the high-stakes test, such as NUEEFL.
Nonetheless, despite the heated nature of the debates, there has
been little empirical evidence for the validity of the NUEEFL test
and its fairness. Specifically, there is no ample evidence of test
fairness among male and female test takers. In the absence of such
evidence, any talk of the fairness of the selection policy would be
doomed to fail.
The present study aims at investigating the validity of a
high-stakes test in general and to considering the role of gender
as a source of bias in the NUEEFL, in particular. Regardless of the
content of the debates over the gender issue, it appears that there
is no evidence on the effect of gender on the performance on the
NUEEFL. If gender asserts a large influence, then it would be a
case of bias and will undermine validity of the test. This is
because gender is not part of the construct measured by the test
and any significant impact by gender is a case of
construct-irrelevant variance. As a part of standard process,
Differential Item Functioning (DIF) analysis is conducted on the
test items, as a main factor in the evaluation of the fairness, and
validity of educational tests.
www.moj-es.net
10
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
In order to investigate the psychometric properties of the
high-stakes test (i.e., NUEEFL), the present
study will address the following research questions: 1. To what
extent do the item responses of NUEEFL form a unidimensional
construct according to the
Rasch measurement model? 2. Is participants gender a source of
DIF in NUEEFL items?
Review of Related Literature Differential Item Functioning (DIF)
Test developers deploy several quality control or statistical
procedures to ensure that the test items
are proper and fair for all examinees (Camilli & Penfield,
1997; Holland & Wainer, 2012; Ramsey, 1993). The statistical
procedure aims at identifying items with different statistical
features across certain groups of examinees. This refers to
differential item functioning (DIF) and such items are said to
function differentially across groups, which is a potential
indicator of item bias (Sireci & Rios, 2013, p. 170).
According to Geranpayeh and Kunnan (2007) irrespective of
considering the fairness issue in the
design-development-administration-scoring cycle, still many
problems are found in this procedure. Geranpayeh and Kunnan (2007)
maintained there are two approaches for solving these problems. One
approach is to develop a pilot group in order to examine test
scores. If the test has been already conducted, a large sample is
available to examine test scores and to investigate items and
functions. If it is identified that they act differently, the
source of this difference is known as Differential Item Functioning
(DIF).
To characterize the definition of DIF, Wiberg (2007) indicated
that identifying problematic items via item analysis plays a key
role in a test. It is maintained that item analysis includes using
statistical techniques to examine the test takers performance on
the items (Wiberg, 2007, p. 1) and one of the crucial parts in the
item analysis is to detect differential item functioning. The DIF
technique is still a very useful method for identifying potential
problem items (Angoff, 1993).
Differential item functioning occurs when an items properties in
one group are different from the items properties in another group
(Furr & Bacharach, 2007, p. 331). To highlight this point, Furr
and Bacharach (2007) specified via an example; DIF exists when a
particular item has different levels of difficulty for males and
females. Put another way, the incidence of differential item
functioning means that a male and a female who have the same trait
or ability level have different probabilities of answering the item
correctly. It is concluded that the presence of DIF between groups
shows that the groups cannot be meaningfully compared on the item
(Furr & Bacharach, 2007).
DIF procedures are used to determine whether the individual
items on a test function in the same way for two or more groups of
examinees, usually defined by racial/ethnic background, sex,
age/experience, or handicapping condition (Scheuneman &
Bleistein, 1989, pp. 255-256). A plethora of studies categorized
DIF detection techniques. To date, many DIF analysis techniques
have been proposed. McNamara and Roever (2006, p. 93) classified
methods for detecting DIF into four broad categories; 1). Analyses
based on item difficulty: These approaches compare item difficulty
estimates. 2). Nonparametric approaches: These procedures use
contingency tables, chi-square, and odds ratios. 3).
Item-response-theory-based approaches which include 1, 2, and
3-parameter IRT analyses. 4). Other approaches. These include
logistic regression, which also employs a model comparison method,
as well as generalizability theory and multifaceted measurement,
which are less commonly used in classic DIF studies.
A large range of possible techniques is available; however, only
a limited number are currently used. The following section attempts
to consider Item Response Theory (IRT)-based models, specifically
the Rasch model as an applicable and germane method to present
research.
The Rasch Model IRT is an extension of classical testing theory
with mathematical roots which deeply penetrated in
psychology and the mathematical basis of IRT has been embedded
in the psychological measurement (Ostini & Nering, 2006). Some
controversial issues, however, exist in defining the concept of
measurement in human science and psychology. Rasch model is
mathematically equivalent to the one-parameter logistic (1PL) IRT
model, but they developed separately (DeMars, 2010). Controversy
surrounds the Rasch model; some specialists believed that the Rasch
model and IRT models are structurally different and are used very
differently. It is claimed that IRT models are used to describe and
fit data; when fit is poor, the model is adapted or discarded in
favor of another model and, in contrast, the Rasch model is more
prescriptive. The
www.moj-es.net
11
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
data are required to fit the model and when they do not, items
that show misfit are discarded until a satisfactory fit is obtained
(Zand Scholten, 2011, p. 39).
The Rasch model involves model-based measurement in which trait
level estimates depend on both the persons responses and on the
properties of the items that were administered (Embretson &
Reise, 2000, p. 13). Furthermore, the test items should not act
differently for any specific subgroups of the participants. If an
item behaves differently for particular groups, then the validity
of the measure for the certain construct decreases; as it is
considered as a threat to the test fairness. The Rasch model
approach permits investigation of the biased items toward different
subgroups and to inspect the construct irrelevant factors (i.e.,
gender, ethnicity, and academic background) via calculating
Differential Item Functioning (DIF) measures.
Besides that, the Rasch model assumptions include
unidimensionality and local independence. A unidimensional test
consists of items that refer to only one dimension; as DeMars
(2010) asserted whenever only a single score is reported for a
test, there is an implicit assumption that the items share a common
primary construct (p. 38). Wale (2013) mentioned that the
assumption of unidimensionality requires the items function in
unison and all non-random variance in the data can be accounted for
by person ability and item difficulty (p. 56). Generally,
unidimensionality indicates whether the items makes a single latent
trait () (DeMars, 2010).
One aspect regarding unidimensionality deserves caution.
Sometimes, responses to test items can be mathematically
unidimensional while the items measure what educators and
psychologist would conceptualize as two different constructs. For
example, test items may gauge both test-taking speed and knowledge
(DeMars, 2010).
Another assumption of the Rasch model is local independence.
Local independence is the probability of a test taker responding
correctly to a certain item is not dependent on previous responses
or the answers given by other individuals to the same item (Wale,
2013, p. 56). Unidimensionality can be checked via model fit
statistics. Besides, unidimensionality and local independence are
estimated using fit model statistics; to say a person or an item
may be misfitting means the extent to which an intended person and
item does not act as the Rasch model would predict.
METHODS
Participants The participants of the present study (N = 5000)
were selected from among the pool of examinees
from a population of 20,000 who had taken a recent version of
the NUEEFL test in 2015. The participants were randomly selected
from the two gender groups (i.e., males and females). The female
group included 3335 persons of the total participants and the rest
of 1665 examinees were male. The academic background and the age of
the participants were not considered in this study.
Instrumentation The National University Entrance Examination for
Foreign Languages (NUEEFL) has a total of 95 items
of which 25 are general questions and 70 items come under six
subtests: a) Grammar (10 items), b) Vocabulary (15 items), c)
Sentence Structure (5 items), d) Language Functions (10 items), e)
Cloze Test (15 items), f) Reading Comprehension (15 items).
The NUEEFL test is annually administered to more than 100,000
university applicants to attempt to find the B.A degree in
governmental university, specifically in the field of foreign
languages. The questions all are in multiple-choice format and are
dichotomously scored. The test is time restricted with a dedicated
time of 105 minutes. Generally, as a rule in NUEEFL test, guessing
is not allowed and the test has included negative score for the
wrong responses. It means that a total of three wrong answers will
expurgate a correct answer.
The latest version of Winsteps software Version 3.92.1 updated
in February 2016 was employed for the data analysis (Linacre,
2016a, b). Winsteps constructs Rasch measures from simple data sets
(i.e., usually of persons and items) and applies the dichotomous
Rasch model. Pearson-test reliability and item reliability of
NUEEFL test were excellent (Pearson r = 0.93 and item r = 1).
www.moj-es.net
12
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Procedure The NUEEFL is administered annually to a large number
of test takers all across Iran. The present
study focused on the one aspect of test validity which was
assessed through the implementation of the Rasch model. To
investigate DIF analysis and apply the Rasch model, the statistical
and mathematical assumptions must be met.
Data Analysis The psychometric properties of the items were
estimated using the Winsteps software (Linacre,
2016b). Since the dataset was dichotomous the data were analyzed
implementing the Joint Maximum Likelihood Estimation (JMLE) method
for estimating the Rasch parameters. In the JMLE formula, the
estimate of the Rasch parameter happens when the observed raw score
for the parameter matches the expected raw score.
The data-model fit estimated through employing the infit and
outfit mean-square values to identify misfit and good-fit items.
When it is said a person or an item may be misfitting, it denotes
that an intended person and item does not act as the Rasch model
would predict (Boone et. al., 2014). The fit estimation checks for
the model mis-specifications that can be evaluated in the fit
between the model and the data (DeMars, 2010). There are two
different fit statistics for persons or items; they are called the
weighted (infit) which weights the square residual by the variance
of item, while the unweighted (outfit) gives the residual the same
weight (i.e., 1) (Wale, 2013, p. 57). The normal range of
acceptable fit for both statistics is between 0.70 and 1.3 (Bond
& Fox, 2007; Liu, 2010).
Furthermore, like many IRT models, the Rasch model rests on two
basic assumptions: unidimensionality and local independence. The
unidimensionality assumption requires that there is only one
underlying construct measured by the set of items included in the
test. That is, the test measures only one factor. The local
independence assumption requires that an examinees response to an
item does not influence his or her response to any other item.
Hence, the items must not give a clue to the correct response for
another item.
Unidimensionality was checked through Principal Component
Analysis (PCA) in Winsteps. What is required is the fact that there
must one dominant factor explaining the shared line of covariance
among the items (Hambleton, Swaminathan, & Rogers, 1991).
Hence, unidimensionality will hold if the first extracted factor
explains a much higher amount of the total variance than that
explained by the secondary dimensions. As mentioned before,
multiple methods for assessing unidimensionality exist, including
the data-model fit statistics. However, studies indicated that
these statistics lack the sensitivity required to detect
multidimensionality. Hence, it is logical to use Principal
Components Analysis (PCA) on the raw data and residuals, in
addition to checking the data-model fit.
One approach, first proposed by Smith (2002), for assessing
unidimensionality within the Rasch model framework, is the
Principal Component Analysis (PCA) based method. This approach aims
at assessing whether the items are unidimensional enough as to be
treated in practice (Smith, 2002). Principal Component Analysis has
the advantage of compressing the data, once patterns have been
found in the data. It reduces the number of dimensions, without
losing too much of information (Smith, 2002). PCA determines
whether the set of items represents a single construct or not. The
analysis of dimensionality, based on Smiths approach, involves a
two-step process.
First, the measurement dimension of the scale was estimated
using the Rasch model. The variance associated with this
measurement dimension was extracted from the item-response data by
computing standardized residuals: (observed - expected)/ (model
standard error). Second, a principal component analysis of the
standardized residuals was used to determine whether substantial
subdimensions existed within the items. If the items measure a
single latent dimension as estimated by the Rasch model, then the
remaining residual variance should reflect random variation.
McCreary et al., 2013, p. 6
For determining the unidimensionality of questions in the PCA
method, conventional criteria were used for judging
unidimensionality (Linacre, 2006). Regarding this, Reckase (1979)
suggested the following criteria for unidimensionality: a). if the
amount of variance explained by measures be > 20%, b). the
www.moj-es.net
13
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
unexplained variance of the eigenvalue for the first contrast
(size) < 3.0 and unexplained variance explained by first
contrast < 5% is good (Linacre, 2006, p. 272).
According to Smiths approach, at first, the parameters for all
questions were estimated, which is called level A in the present
research. In the PCA approach, item residuals with loadings of +0.3
or more and 0.3 or less are taken as potential representatives of
subdimensions (Hagell, 2014). And in the second round of data
analysis, it is attempted to detect the questions with the outfit
MNSQ statistic value larger than 1.3. This phase is called level B
in which the parameters of the questions were separately estimated.
Then, the difference in the difficulty parameter of the questions
which were obtained in level B, from the difficulty parameter in
level A was estimated. Accordingly, the mean scores of the
differences was calculated. Technically, the estimation is called
constant correction.
In the next step, the participants ability parameter were
calculated once regarding the entire test (Level A), and once the
items calculated separately (Level B). Afterwards, the constant
correction value which had been previously calculated was added to
the ability parameter (i.e., the Level B). Finally, the difference
between the individuals ability level in the entire test and in the
separate items in Level B were calculated through a series of
independent t-test for determining the statistical significance and
to compare the two estimates on a person-by-person basis in order
to determine the proportion of instances in which the two item sets
yield different person measures (Hagell, 2014, p. 460).
In order to determine the significance of the t-test, the error
level of type one ( = 0.05) has been modulated. Smith (2002)
indicated that if the level of significance of t-test exceeds 5%,
the local independency and unidimensionality will be violated.
Moreover, the DIF analysis was examined to test the invariance
of measurement. The test developers deploy several quality control
or statistical procedures to ensure that the test items are proper
and fair for all examinees (Camilli & Penfield, 1997; Holland
& Wainer, 2012; Ramsey, 1993).
According to Angoff (1993) an item which shows DIF has different
statistical properties in different groups when monitoring for
differences in the abilities of groups. DIF is highlighted as an
unexpected difference between two groups after matching to measure
the underlying ability in the intended item (Camilli, 2006; Wiberg,
2007). Besides, the essential component of DIF analysis in the
Rasch model is to compare the item difficulties obtained from the
two samples. If the difference in the difficulty estimates is
large, then measurement invariance fails and DIF has happened. In
the present study the DIF analysis was carried out to test whether
the items functioned differently across gender groups.
RESULTS
Data-model Fit Estimation
The Winsteps software normally assessed the fit of the model
through obtained statistics indicators of mean-square fit values
(MNSQs) and the standardized Z values (ZSTDs). The values in the
range of MNSQs are considered from zero to infinite (0- ) and the
expected value is 1. Values above 1 show a deviation from the
unidimensionality, and values less than 1 indicate the overfit in
the response patterns with the data-model. The overfit in the model
implies the existence of dependency among responses or items.
It is worth mentioning that this statistical index is very
sensitive to the sample size. For the sample size with less than 90
people, the model fits with any types of data model, whereas the
model does not fit with samples consisting more than 900
people.
Therefore, due to the large sample size in this study and
keeping with the valued guidance provided by Linacre (2012), the
data-model fit was displayed through MNSQs. The Rasch model offers
two indicators of misfit: the infit and outfit mean square indices.
Infit is sensitive to unexpected responses to items near the
persons ability level and outfit discusses difference between
observed and expected responses regardless of how far away the item
endorsability is from the persons ability (McCreary et al., 2013,
p. 7). The MNSQs estimates and reports both outfit and infit MNSQs
for analyzing the fit of the model.
For both indicators, values between 0.701.3 are considered as
acceptable or so called good fit values.
www.moj-es.net
14
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Values less than 0.70 indicate outfit, whereas values above 1.3
are a sign of infit. Furthermore, Linacre favors outfit-MNSQs over
infit-MNSQs. Hence, in the present research in order to assess data
model fit it is decided to consider outfit-MNSQs as a criterion for
the outcome interpretations. And, the acceptable values for this
index are in the range of 0.70 to 1.3.
In analyzing the model fit estimation, it is required to
eliminate the participants with total score of zero. The data were
screened for outliers. Besides, the fit indices should be reported
for the item calibration. The difficulty estimates for the items,
standard errors of item difficulty of estimates, and the
infit-MNSQs and the outfit-MNSQs indices are shown in Table 1.
Note that in Table 1 due to space restriction, only the
estimation of difficulty parameter and model fit estimations of
misfit items are illustrated in descending order (from the most
difficult to the least difficult).
Table 1. Item Statistics for Fit Model Estimate and Difficulty
Parameter in the Entire Test (Descending Order)
Item Entry Number Total Score
Total Count Measure
Model S.E.
Infit MNSQ
infit ZSTD
Outfit MNSQ
Outfit ZSTD
Q155 80 158 5000 2.45 0.08 1.07 1 1.82 4.8 Q126 51 191 5000 2.23
0.08 1.08 1.2 2.05 6.3 Q137 62 265 5000 1.84 0.07 1.04 0.8 1.32 2.6
Q105 30 285 5000 1.76 0.07 1.19 3.7 3.01 9.9 Q118 43 314 5000 1.64
0.06 1.18 3.6 1.98 7.3 Q166 91 330 5000 1.57 0.06 1.05 1.2 1.49 4.2
Q101 26 335 5000 1.56 0.06 1.26 5.3 3.3 9.9 Q103 28 363 5000 1.46
0.06 1.28 6.1 2.99 9.9 Q158 83 367 5000 1.44 0.06 1.14 3.1 1.46 4.2
Q111 36 392 5000 1.36 0.06 1.12 2.9 1.87 7.4 Q115 40 411 5000 1.3
0.06 1.14 3.5 2.02 8.6 Q133 58 411 5000 1.3 0.06 1.1 2.5 1.77 6.8
Q121 46 459 5000 1.15 0.05 1.38 9.2 2.43 9.9 Q167 92 462 5000 1.14
0.05 0.83 -4.9 0.64 -4.7 Q109 34 463 5000 1.14 0.05 1.24 6 2.03 9.2
Q128 53 496 5000 1.05 0.05 1.2 5.3 1.44 4.7 Q122 47 600 5000 0.79
0.05 1.13 4.2 1.46 5.4 Q156 81 732 5000 0.5 0.04 1.2 6.9 1.41 5.4
Q99 24 821 5000 0.33 0.04 0.83 -7.1 0.69 -5.7 Q84 9 860 5000 0.26
0.04 1.28 9.9 1.6 8.5 Q153 78 874 5000 0.24 0.04 0.78 -9.4 0.59
-8.1 Q149 74 1046 5000 -0.05 0.04 0.75 -9.9 0.57 -9.9 Q108 33 1079
5000 -0.1 0.04 1.17 7.6 1.33 6.1 Q160 85 1151 5000 -0.21 0.04 0.8
-9.9 0.67 -8.1 Q91 16 2076 5000 -1.37 0.03 0.76 -9.9 0.67 -9.9 Q79
4 2513 5000 -1.85 0.03 1.2 9.9 1.36 9.9 Mean 1170.9 5000 0 0.04 1
-0.7 1.14 0.1 P.SD 790.5 0 1.21 0.01 0.13 5.4 0.5 5.3 The first
column of Table 1 is the Item Number. The second column shows the
Entry Number which
provides the order of entering the data in each row. The next
column provides Total Score which is total number of correct
answers and Total Count is the total number of participants or
responses. The fifth column, the measure, reports the difficulty
estimates for the items. The range value of difficulty parameter is
from
www.moj-es.net
15
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
2.45 to -2.99, with mean score of 0, and Standard Deviation (SD)
of 1.21. The descending management of item statistics in Table 1 is
helpful to arrange the most difficult item Q.155 (Measure = 2.45)
and the least difficult item Item Q.87 (Measure = -2.99). Column
six provides the Standard Error (SE) of item difficulty
estimates.
In the last four columns the infit and outfit statistics are
presented. In the following tables the infit-MNSQs values were
provided, however; in estimation of data model fit the values of
outfit-MNSQs were merely used. As explained before, the acceptable
range for both infit and outfit MNSQs is between 0.70 and 1.33. In
this study, the outfit- MNSQs indices are all within the acceptable
range. However, it appears 26 items were not located in the
acceptable range. In Table 1 the range value of the outfit-MNSQs
varies from 0.57 to 3.33 which means that some items presented in
Table 1 do not fit the model. The investigation of item statistics
of outfit-MNSQs reveals that 26 items (27% of items) are not
fitted.
Figure 1. The ICC Curve for Misfitting Item, Item 105.
Figure 1 shows the Item Characteristic Curve (ICC) for Item 105.
The red curve is the expected ICC. It would be gained if the data
fitted the Rasch model. The blue curve is the observed or empirical
ICC. The grey line in the outskirt of the red curve is the
confidence interval. The confidence intervals are constructed from
an estimate and its standard error. The data in present study
employed from a large data set, it is evident that the standard
error would be very small. The confidence interval become wider if
the output of data analysis contains the large standard error.
Figure 1 shows that the empirical ICC for mis-fitting Item 105
has a large deviation from the expected ICC. This fact is reflected
in Item 105 with the large outfit-MNSQs value of 3.01. On the
contrary, for instance, Figure 2 demonstrates the empirical ICC for
good-fitting item (i.e., Item 142). The out-fit MNSQs value (i.e.,
0.94) is within the acceptable range.
www.moj-es.net
16
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Figure 2. The ICC Curve for Good-fit Item, Item 142.
The presence of a large number of misfitting items demonstrates
that the data does not fit the model in the NUEEFL. Therefore, the
model and its assumptions may be violated. It is possible that the
Rasch model unidimensionality assumptions also may not attain the
desirable results. Thus, in the next section the results of
unidimensionality and local independence will be reported.
Unidimensionality
The unidimensionality assumption requires that there is only one
underlying construct measurement by the set of items included in
the test. That is, the test measures only one factor. There are
multiple methods for assessing unidimensionality, including the
data-model fit statistics. However, studies indicated that these
statistics do not have the ample sensitivity required to detect
multidimensionality. Hence, it is logical to use a Principal
Component Analysis (PCA) on the raw data and residuals, in addition
to checking the data-model fit. In the current research, the
unidimensionality of the test and its items were checked through a
Principal Components Analysis (PCA) of residuals and a t-test.
In order to assess dimensionality, PCA of the Rasch residuals
was performed. The variance of the measurement dimension is 34.8%
with 12.7% of raw variance explained by persons and 22.1% raw
variance explained by items. The results showed that it is larger
than the requirement of 20%, demonstrating a unidimentional trait
of the data (Reckase, 1979).
The first, second, third, fourth, and fifth unexplained variance
accounted for eigenvalues are 3.4, 2.5, 2.2, 1.9, and 1.7 which
were good by referring the criteria. The results of the data
analysis suggested that the unidimensionality is hold across the
whole test (See Table 2 & 3).
www.moj-es.net
17
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Table 2. PCA Analysis
Table 3. Contrast in Eigenvalue Units
Expected Observed Eigenvalue Contrast in Eigenvalue units 3.6%
2.4% 3.4662 Unexplained variance in 1st contrast 2.7% 1.7% 2.5397
Unexplained variance in 2nd contrast 2.4% 1.5% 2.2560 Unexplained
variance in 3rd contrast 2.0% 1.3% 1.9438 Unexplained variance in
4th contrast 1.8% 1.2% 1.7076 Unexplained variance in 5th contrast
100.0% 100.0% 95.0000 Raw unexplained variance (total)
Local independence was examined through checking all the
abilities in order to identify whether the
responses to items could be independent of each other (Pae,
2011). As for local independence, in assessing the t-test
statistics the outfit-MNSQs was examined. The results showed that
20 items for this statistic exceeded 1.3. These 20 items are
considered as those which confirm the unidimensionality assumption
in data analysis.
The steps of t-test calculation were conducted. Note that the
measurement of local independence was based on Smiths approach
(2002). The total sum of difference between difficulty parameter of
these 20 items gained from level A and B is equal with -1.12. The
constant correction value is -0.056 which is obtained through
dividing -1.12 by the number of items (20 items). By adding the
constant correction value to the ability parameter of participants
within questions of level B. Then, it is attempted to examine the
significance level of t-test and to modify the significance
level.
The Students t-statistics on 20 items revealed that there were
no significant different among t-test results. Thus, it is
concluded that the local independence assumption is strongly
accepted in the entire NUEEFL test. To sum, regarding the results
both Rasch assumptions (i.e., unidimensionality and local
independence) are hold in the whole test.
Differential Item Functioning The next step in data analysis was
the DIF analysis. The ability and difficulty estimates in the
Rasch
model are assumed to be invariant. The statistical procedure
aims at identifying items with different statistical features
across certain groups of examinees.
In this study, DIF analysis was investigated for the gender
groups and for the NUEFFL items. For DIF Analysis in a Rasch
context, both magnitude of the difference in logit units between
the groups and statistical significance of the difference should be
considered (Linacre, 2016a). The magnitude of the DIF value should
be at least 0.5 logits, indicating the comparison between
differences in difficulty of the items for one group to the
difficulty level of the same items for the other group (Linacre,
2016a).
In this stage of this study, DIF analysis was tested between two
groups of males and females. In order to examine the invariance,
the difference between the DIF analysis of two groups of males and
females through testing the t-test of the statistical significance
of the data was investigated. For statistically significant DIF,
the probability of such difference (0.5 logits or larger),
occurring as a random accident, should be 0.05. This indicates that
the probability of such difference happens when there is no
systematic item bias in the test items (Linacre, 2016a).
Beside, considering that statistical significance tests are
affected by sample size and due to the large size of the study
groups in present research, it needs at least 0.5 logits for DIF to
be noticeable. For instance,
Expected Observed Eigenvalue Variance in Eigenvalue units 33.7%
34.8% 50.7068 Raw variance explained by measures 12.3% 12.7%
18.5454 Raw variance explained by person 21.4% 22.1% 32.1614 Raw
variance explained by items 66.3% 65.2% 95.0000 Raw unexplained
variance 100.0% 100.0% 145.7068 Total raw variance in
observations
www.moj-es.net
18
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
if the difficulty of an item in both groups of males and females
has 0.5 logits difference, that specific item will be considered as
a DIF item. Note that, due to space limitation, only DIF-flagged
items appear in Table 4.
Table 4. DIF-flagged items in the NUEEFL test
Item Number
Person Class
DIF Measure
Person Class
DIF Measure
DIF Contrast
Rasch-Welch
t df Prob
Q76 M -0.87 F -0.67 -0.2 -2.75 INF 0.0059
Q77 M -1.95 F -2.11 0.15 2.19 INF 0.0289
Q80 M -2.5 F -2.08 -0.42 -5.85 INF 0.0000
Q85 M -0.16 F 0 -0.16 -1.96 INF 0.0498
Q86 M -1.17 F -1.02 -0.15 -2.05 INF 0.0400
Q90 M 0.29 F 0.81 -0.52 -5.6 INF 0.0000
Q91 M -1.48 F -1.31 -0.17 -2.47 INF 0.0137
Q94 M -0.99 F -0.81 -0.18 -2.5 INF 0.0123
Q95 M -1.08 F -0.72 -0.35 -4.83 INF 0.0000
Q97 M -0.41 F -0.12 -0.29 -3.59 INF 0.0003
Q99 M 0.14 F 0.45 -0.31 -3.52 INF 0.0004
Q103 M 1.62 F 1.36 0.26 2.1 INF 0.0362
Q104 M 0.55 F 0.35 0.21 2.22 INF 0.0263
Q105 M 2.01 F 1.61 0.4 2.89 INF 0.0039
Q106 M -0.46 F -0.09 -0.37 -4.66 INF 0.0000
Q107 M -0.4 F -0.63 0.23 2.93 INF 0.0034
Q108 M 0.03 F -0.17 0.2 2.42 INF 0.0157
Q111 M 1.62 F 1.21 0.41 3.35 INF 0.0008
Q113 M -0.48 F -0.07 -0.41 -5.15 INF 0.0000
Q117 M -1.19 F -0.98 -0.21 -2.93 INF 0.0034
Q119 M 1.31 F 1.67 -0.36 -2.99 INF 0.0028
Q121 M 1.52 F 0.96 0.55 4.77 INF 0.0000
Q122 M 0.96 F 0.7 0.26 2.58 INF 0.0100
Q125 M 0.36 F 0.81 -0.44 -4.71 INF 0.0000
www.moj-es.net
19
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
Item Number
Person Class
DIF Measure
Person Class
DIF Measure
DIF Contrast
Rasch-Welch
t df Prob
Q128 M 1.2 F 0.96 0.24 2.2 INF 0.0279
Q130 M 0.39 F 0.18 0.21 2.37 INF 0.0179
Q131 M -0.18 F -0.41 0.22 2.78 INF 0.0054
Q132 M -1.38 F -1.7 0.32 4.55 INF 0.0000
Q135 M -0.96 F -1.42 0.46 6.43 INF 0.0000
Q138 M -0.86 F -1.07 0.21 2.84 INF 0.0046
Q141 M -1.81 F -2.02 0.21 2.95 INF 0.0032
Q142 M 0.08 F -0.12 0.2 2.35 INF 0.0187
Q143 M -0.49 F -0.3 -0.19 -2.46 INF 0.0140
Q145 M -0.9 F -1.14 0.24 3.3 INF 0.0010
Q147 M 1.74 F 1.33 0.41 3.28 INF 0.0011
Q157 M -0.04 F 0.24 -0.28 -3.3 INF 0.0010
Q159 M 0.47 F 0.97 -0.5 -5.12 INF 0.0000
Q163 M 0.05 F -0.18 0.22 2.69 INF 0.0072
Q167 M 0.96 F 1.27 -0.3 -2.8 INF 0.0052
Q168 M -0.42 F -0.89 0.47 6.22 INF 0.0000
Note. M = Male; F = Female; INF = Infinity
The DIF analysis between groups of male and female was
investigated. The results of this analysis are shown in Table 4.
The results show that among 95 items, 40 items exhibit DIF.
The difficulty level of items between males and females was
variant. Hence, it is concluded that the invariability of questions
in gender group is not accepted. The null hypothesis which stated
the participants gender is not a source of DIF in NUEEFL is
rejected. Given significance DIF within the Rasch model in gender
group, the NUEEFL appeared not to be a DIF-free person estimates
test. Hence, it is concluded that there is significant difference
between males and female in answering the NUEEFL test. And, the
NUEEFL test is not fair to all male and female participants.
DISCUSSION AND CONCLUSIONS
The present study aimed at investigating the interaction of
person abilities and item difficulties of the high-stakes test of
university entrance exam in Iran (i.e., NUEEFL). In particular,
results from the Rasch model and DIF analysis were compared to see
whether evidence of differential functioning would be found in data
analysis.
www.moj-es.net
20
-
Malaysian Online Journal of Educational Sciences 2017 (Volume5 -
Issue 1 )
The descriptive statistics revealed that there was significant
difference in the overall test. Hence, the
hypothesis that the data fit the Rasch model was not supported.
The results of data-model fit revealed that 26 items (i.e., 155,
126, 137, 105, 118, 166, 101, 103, 158, 111, 115, 133, 121, 167,
109, 128, 122, 156, 99, 84, 153, 149, 108, 160, 91, and 79) out of
total 95 items were not located in acceptable range value of 0.70
to 1.30. There are many misfitting items and items were not fitted
with the model.
A concern about the dimensionality of the NUEEFL test suggests
that a calibration of test of Rasch model in Winsteps software
revealed that there are many misfitting items. The dimensionality
was detected through the Principal Component Analysis (PCA) on the
raw data and residuals. The amount of variance explained by the
different components in the data was 34.8% (eigenvalue 50.70) which
is larger than 20% as to be indicative of unidimensionality. The
results of the data analysis suggested that the unidimensionality
is hold across the whole test.
In the case of local independence, a series of t-test was
performed. The result of outfit-MNSQs showed that 20 items are
larger than the intended criteria (i.e., 1.3). It is concluded that
the local independence assumption is strongly accepted in the
entire NUEEFL test.
DIF analysis confirmed a different probability of endorsing the
test items across the gender groups. Based on the DIF results, it
is interpretable that out of the 95 items, 40 items displayed
DIF-flagged items. This suggests that NUEEFL test scores are not
free of construct-irrelevant variance. Hence, it does not support
the argument for the construct validity.
Additionally, there is an ongoing interest in comparing
cultural, ethnic, or gender groups. DIF studies are essential in
testing programs with high stakes. Furthermore, possible gender
and/or ethnicity bias could negatively impact one or more groups in
a construct-irrelevant manner. In fact, the test administrators
attempt in making a perfectly fair testing battery; however the
dearth of research on NUEEFL test raised questions regarding the
fairness of this national test.
The NUEEFL, which is given to thousands of individuals annually,
is used as a gate-keeping test for entering higher education in
Iran. In line with the main purpose of this research, DIF analysis
across gender groups was investigated. DIF analysis rejected a
similar probability of endorsing the test items across the gender
groups. The results of the study indicated that 40 items out of the
95 items of NUEEFL test exhibited DIF. This suggests that NUEEFL
test scores are not free of construct-irrelevant variance.
It is worth mentioning that the results of the present study are
not consistent with Karamis (2015) study from the aspect of
dimensionality of NUEEFL, in which the multidimensionality in the
whole test and among sub-tests had been proven.
According to Camilli (2006) DIF analysis mainly focuses on the
performance of two or more different groups. Therefore, such
analysis is unable to disclose the existence of bias against
different individuals. Moreover, this study directs to this point
that due to the asses