Naslov
Influence of item format on psychometric properties and latent structure of composite tests

Autori
Ljubotina, Damir

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
20. Dani Ramira i Zorana Bujasa, knjiga sažetaka / Čorkalo Biruški, Dinka ; Vukasović, Tena - Zagreb : Naklada Slap, 2011, 41-41

ISBN
978-953-175-391-3

Skup
20. Dani Ramira i Zorana Bujasa

Mjesto i datum
Zagreb, Hrvatska, 07.04.2011. - 09.04.2011

Vrsta sudjelovanja
Pozvano predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Item response theory; item format; latent structure

Sažetak
Unlike classical test theory (CTT), item response theory pays a great deal of attention to item-analysis, and several specific models have been developed to assess various test and item types. These models also provide a framework for assessing model fit with regard to empirical data. Two basic item types are multiple choice and short-answer. Short answer items introduce a different kind of guessing related error variance, which is not easily accounted for. These two item types differ in difficulty, provoke different processes (recognition vs. recall), and elicit different subject strategies with regard to guessing ; especially when they not sure they know the answer. There are degrees of guessing from blind to nearly complete knowledge, and in light of that the meaning of „guessing” becomes ambiguous. Previous studies have shown that the effect of instructions vary over subjects and create an irrelevant source of individual differences. Consequently, individual differences in guessing strategies complicate test factor structure. Studies have also demonstrated that short-answer tests contain measurement error that is not present in multiple choice tests. With multiple choice tests overall test reliability increases as the function of number of alternatives, however this increase is smaller than that predicted by the blind guessing model. Also, there is a larger reliability gain for low-scoring individuals since they seem to be more prone to guessing. Our goal was to determine how different item types influence test characteristics. To reach that goal we have varied the type items (2, 3, 4, 5, 6 alternatives and short-answer) while controlling for item content, and used two different test instructions. Relevant effects are discussed in the context of 1, 2 and 3-PL model under item response theory framework as well as under classical test theory assumptions.

Izvorni jezik
Engleski

Znanstvena područja
Psihologija

POVEZANOST RADA