Naslov
Latent structure and psychometric properties as a function of test item characteristics

Autori
Ljubotina, Damir

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

Izvornik
18. Dani Ramira i Zorana Bujasa: Program i sažeci priopćenja / Keresteš, Gordana ; Brković, Irma ; Butković, Ana - Zagreb : Odsjek za psihologiju Filozofskog fakulteta Sveučilišta u Zagrebu, 2007, 50-50

ISBN
978-953-175-302-9

Skup
18. Dani Ramira i Zorana Bujasa

Mjesto i datum
Zagreb, Hrvatska, 13.12.2007. - 15.12.2007

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Domaća recenzija

Ključne riječi
item response tehory; item characteristics; latent structure

Sažetak
Th e goal of the study was to examine the infl uence of some item characteristics on the basic test psychometric properties such as reliability, homogeneity, criterion validity and factor structure. Most of the classic parameter assessment procedures are open to problems of variance reduction which can be the result of limited item values (e.g. binary items), so when examining factor structure hyperfactorization problems are observed. A simulation study with a factorial 3 x 3 x 3 experimental design was conducted. Th e infl uence of three main eff ects was examined: (a) possible range of item values (which infl uences variance reduction), (b) items diffi culty for the specifi ed levels of ability (which infl uences the shape of the distribution), and (c) items discriminant power. Th e known true parameter values were compared with the estimates obtained in diff erent experimental conditions. Internal consistency reliability measures, number of signifi cant factors, average inter-item correlation, and measures of discriminant and criterion validity were used as dependent variables. Results showed a signifi cant infl uence of the eff ects on the dependent variables, and showed that variance reduction and item distribution asymmetry have a signifi cant eff ect on the estimates of items’ psychometric characteristics and their factor structure. Possible procedures for the correction of the observed eff ects are discussed. It seems that the best possible direction to follow is to compare the observed results with the data from simulated equivalent data matrices. Th e author discusses these eff ects in the context of 2-parameter logistic model of item response theory and under the assumptions of the classical test theory.

Izvorni jezik
Engleski

Znanstvena područja
Psihologija

POVEZANOST RADA