Progress Test (PT) is a form of assessment that simultaneously measures ability levels of all students in a certain educational program and their progress over time by providing them with same questions and repeating the process at regular intervals with parallel tests. Our objective was to generate an item bank for the PT and to examine the possible fit of CAT for PT application. This study is a descriptive study. 1206 medical students participated. During the analysis of the psychometric properties of PT item bank, “the Rasch model for dichotomous items was used”. Several CAT simulations were performed by applying various stopping rules of different standard errors. CAT simulation estimates were compared with the estimates generated from the original calibration of the Rasch model where all items were included. After Rasch analysis, a unidimensional PT item bank consisting of 103 items was obtained. The item bank reliability was calculated as 0.77 with Person Separation Index (PSI) and Kuder-Richardson Formula 20 (KR-20). A high correlation between θ estimations obtained from paper-and-pencil (θRM) and CAT applications (θCAT) was detected for simulation conditions ([N(0,1)] and [N(0,3)]) at the end of our analysis. In CAT, estimation can be made with an average of 14 questions (reduced 86,4%) and 17 questions (reduced 83,4%) [for N(0,1) and [N(0,3) respectively] with reliability of 0,75. This study reveals that it is possible to develop an appropriate item bank for the PT, and the difficulty of administering large number of items in PT can be scaled down by incorporating CAT application.
Computer Adaptive Test Rasch Models Progress Test Medical Education Item Bank Development
Progress
Test (PT) is a form of assessment that simultaneously measures ability levels
of all students in a certain educational program and their progress over time
by providing them with same questions and repeating the process at regular
intervals with parallel tests. Our objective was to generate an item bank for
the PT and to examine the possible fit of CAT for PT application. This study is
a descriptive study. 1206 medical students participated. During the analysis of
the psychometric properties of PT item bank, “the Rasch model for dichotomous
items was used”. Several CAT simulations were performed by applying various
stopping rules of different standard errors. CAT simulation estimates were
compared with the estimates generated from the original calibration of the
Rasch model where all items were included. After Rasch analysis, a
unidimensional PT item bank consisting of 103 items was obtained. The item bank
reliability was calculated as 0.77 with Person Separation Index (PSI) and
Kuder-Richardson Formula 20 (KR-20). A high correlation between θ estimations
obtained from paper-and-pencil (θRM) and CAT applications (θCAT) was
detected for simulation conditions ([N(0,1)] and [N(0,3)]) at the end of our
analysis. In CAT, estimation can be made with an average of 14 questions
(reduced 86,4%) and 17 questions (reduced 83,4%) [for N(0,1) and [N(0,3)
respectively] with reliability of 0,75. This study reveals that it is possible
to develop an appropriate item bank for the PT, and the difficulty of
administering large number of items in PT can be scaled down by incorporating
CAT application.
Computer Adaptive Test Rasch Models Progress Test Medical Education Item Bank Development
Birincil Dil | İngilizce |
---|---|
Konular | Eğitim Üzerine Çalışmalar |
Bölüm | Makaleler |
Yazarlar | |
Yayımlanma Tarihi | 5 Ocak 2020 |
Gönderilme Tarihi | 21 Ekim 2019 |
Yayımlandığı Sayı | Yıl 2019 Cilt: 6 Sayı: 4 |