Abstract

NUNES, Carlos Henrique Sancineto da Silva  and  PRIMI, Ricardo. Impact of the sample size in the item and subject's parameters estimates under item response theory. Aval. psicol. [online]. 2005, vol.4, n.2, pp. 141-153. ISSN 1677-0471.

Item Response Theory (IRT) has been considered an important development for the modern psychometrics because of its several advantages compared to Classic Test Theory (CTT), such as: the virtual invariance of item parameters in respect to the sample used in their estimation, more reliable and interpretable identification of person`s ability and more efficient procedures for test equating. Nevertheless, there are discussions in respect to the minimal number of subjects in studies using IRT. The aim of the study was to investigate the effect of sample size in the fluctuations of item and person parameters. Nine samples with different sizes were assembled from a database of 44.000 answers from students of Bahia State to an educational exam in mathematics. Results indicated that item and person parameters can be adequately estimated from samples starting form 200 subjects. Smaller sample size produces greater instability with the three-parameter model.

Keywords : Item Response Theory; Sample Size; Statistics and Methodology.

        · abstract in Portuguese     · text in Portuguese     · Portuguese ( pdf )