An empirical evaluation and comparison of Classical Test Theory and Rasch Model

Abstract

Although it is considered an interdisciplinary subject, theory of measurement is mostly based on the mathematical foundations. In the last
quarter of the 19th century, various psychometric approaches have been developed on the strength of those scientific foundations. Those
theories allow users to analyze and determine relationship among individuals’ latent abilities and their response patterns with respect to surveys,
aptitude
tests,
examinations
or
other
educational
measurement
tools.
Besides,
these
models
of
measurement
are
enhanced
to
estimate

model
parameters on
the strength of features of items. Item-person statistics are used for the educational measurement
and
assessment
purposes
mostly.
In
that
context,
Classical
Test
Theory
(CTT)
and
Rasch
model
are
focused
and
elaborated
to
evaluate
School
Olympiad
Examination.
The
way
how
to
implement
both
theories
and
how
to
interpret
results
of
the
analyses
is
presented
in
the
study.

Comparison
of
both

theories is done. This comparison gives chance to determine similarities and differences between them. Using the IRTPRO program, some
notable features of the items are described and discussed in detail. According to gender, item-person statistics/parameters are determined
and interpreted. Behaviors of Item Characteristics Curves are elaborated with respect to both genders. Reliability of the test is defined with
Cronbach’s alpha in the whole test and with respect to genders. As a conclusion it is suggested that SOE (School Olympiad Examination) has
a high reliability as an educational test, groups of female and male students handled almost all of the items similarly. CTT and Rasch model
give similar information with respect to different genders in general. However, Rasch model gives better and more informative results in the
ability determination and comparison.