Abstract
Dimensionality and test information function. Logistic Item Response models assume data unidimensionality, although this assumption is difficult to obtain with empirical data. In this paper, the influence of unidimensionality violations on test information function was investigated. Sixteen data bases with different degrees of dimensionality were generated, by crossing the number of factors (2/3), with type of items distribution (uniform/no uniform), and factors correlations ( 4 levels). 30 items and 1000 subjects were used. Results show that the test information function appears to be robust to mild violations of unidimensionality, decreasing the robustness when the number of dimensions increases.
