Downloads & Free Reading Options - Results

Seven different techniques for transforming two sets of parameters to a common metric in item response theory were evaluated. These techniques include three techniques that estimate the transformation by equating the first two moments of the distributions of estimated difficulty parameters. Another three techniques are included that estimate the transformation by minimizing sums of squares criteria. The remaining technique (MLE) finds the scale transformation that maximizes the likelihood of observing vectors of item parameter differences. Results indicate that for calibrations with a small number of items with extreme estimated difficulty values all techniques perform satisfactorily for moderate sample sizes (500 subjects) and test lengths (60 items). With a large number of extreme estimated difficulty values the three sums of squares and MLE techniques perform satisfactorily, while the difficulty parameter techniques do not. The MLE technique appears poorly suited for smaller samples (500 subjects) and shorter tests (30 items). Keywords: Item response theory; Linking; Equating; Scale transformation; Common metric.