Taiwo Oluwafemi AJEIGBE
taiaje@oauife.edu.ng, ajetaiolu@gmail.com
Department of Educational Foundations and Counselling
Faculty of Education, Obafemi Awolowo
University, Ile-Ife.
and
Olayinka Ib
ASSESSING UNIDIMENSIONALITY AND ITEM PARAMETER ESTIMATES
OF FOUR DIFFERENT PAPER TYPES OF ENGLISH LANGUAGE
MULTIPLE-CHOICE TESTS USING THREE-PARAMETER MODEL
The study examined unidimensionality, ascertained and compared stability of item
difficulty, discrimination and guessing tendencies across four different paper types of
English Language multiple-choice tests of Distance Learning Centre in Obafemi
Awolowo University. These were with a view of establishing parameter stability of the
English Language Multiple-Choice. The study adopted a causal comparative design
because students’ responses were obtained from the database. The 2449 students who
sat for the first contact examination during the 2015/2016 session were used as sample
size for the study. The instruments used for the study were four different paper types of
English Language for 2015/2016 first contact, consisting of 60 Multiple-choice items
each. The items were calibrated to generate item difficulty, discrimination and guessing
tendency using X-Calibre 4.2 software package. One-way analysis of variance was
used to estimate statistical difference in terms of item difficulty, discrimination and
guessing tendency across the four different paper types. Results obtained showed that
each test paper type is unidimensional in nature. Also, out of the 60 items 25(41.7),
39(65.0) and 38(63.3); 32(53.3), 36(60.0) and 42(70.0); 30(50.0), 37(61.7) and
38(63.3); and 27(45.0), 36(60.0) and 40(66.7) fell under moderate difficulty,
discrimination and acceptable guessing value of 0.00 – 0.25 across the four different
paper types respectively, The results finally showed that there was no statistical
significant difference in item difficulty (F(3, 239)=0.028; p>0.05); discrimination (F(3,
239)=0.212; p>0.05); and guessing (F(3, 239)=0.425; p>0.05) respectively. The study therefore concluded that the assumption of unidimensional was not violated and stable
parameter estimates were recorded across paper types.
Keywords: Unidimensionality, Item, Parameter estimates, English Language, Multiple-
Choice Tests, Different paper types, Three-parameter model
Taiwo Oluwafemi AJEIGBE taiaje@oauife.edu.ng, ajetaiolu@gmail.com Department of Educational Foundations and Counselling Faculty of Education, Obafemi Awolowo University, Ile-Ife. and Olayinka Ib
ASSESSING UNIDIMENSIONALITY AND ITEM PARAMETER ESTIMATES OF FOUR DIFFERENT PAPER TYPES OF ENGLISH LANGUAGE MULTIPLE-CHOICE TESTS USING THREE-PARAMETER MODEL
The study examined unidimensionality, ascertained and compared stability of item difficulty, discrimination and guessing tendencies across four different paper types of English Language multiple-choice tests of Distance Learning Centre in Obafemi Awolowo University. These were with a view of establishing parameter stability of the English Language Multiple-Choice. The study adopted a causal comparative design because students’ responses were obtained from the database. The 2449 students who sat for the first contact examination during the 2015/2016 session were used as sample size for the study. The instruments used for the study were four different paper types of English Language for 2015/2016 first contact, consisting of 60 Multiple-choice items each. The items were calibrated to generate item difficulty, discrimination and guessing tendency using X-Calibre 4.2 software package. One-way analysis of variance was used to estimate statistical difference in terms of item difficulty, discrimination and guessing tendency across the four different paper types. Results obtained showed that each test paper type is unidimensional in nature. Also, out of the 60 items 25(41.7), 39(65.0) and 38(63.3); 32(53.3), 36(60.0) and 42(70.0); 30(50.0), 37(61.7) and 38(63.3); and 27(45.0), 36(60.0) and 40(66.7) fell under moderate difficulty, discrimination and acceptable guessing value of 0.00 – 0.25 across the four different paper types respectively, The results finally showed that there was no statistical significant difference in item difficulty (F(3, 239)=0.028; p>0.05); discrimination (F(3, 239)=0.212; p>0.05); and guessing (F(3, 239)=0.425; p>0.05) respectively. The study therefore concluded that the assumption of unidimensional was not violated and stable parameter estimates were recorded across paper types. Keywords: Unidimensionality, Item, Parameter estimates, English Language, Multiple- Choice Tests, Different paper types, Three-parameter model