ERIC Number: ED189165
Record Type: RIE
Publication Date: 1980-Apr
Reference Count: 0
Some Empirical Evidence for Latent Trait Model Selection.
Hutten, Leah R.
The results of this study suggest that for purposes of estimating ability by latent trait methods, the Rasch model compares favorably with the three-parameter logistic model. Using estimated parameters to make predictions about 25 actual number-correct score distributions with samples of 1,000 cases each, those predicted by the Rasch model fit the observed distribution of number-correct scores as well as the distributions predicted by the three-parameter model. This suggests that the Rasch model is quite robust to violations of its underlying assumptions and can therefore be applied in a wide variety of situations. The single model assumption related to fit was that of unidimensionality. Not only was fit for multidimensional tests quite poor, but it was also difficult to obtain parameter estimates on such tests. Other research indicates that the three-parameter model may be preferable for test development, test equating, and studies of item bias, but for estimating ability, this study shows that the Rasch model produces estimates of ability that rival those produced by the three-parameter model. Further, the study demonstrates that Rasch parameter estimates are obtained at costs significantly reduced rom from those of the three-parameter estimates. (Author/CTM)
Publication Type: Speeches/Meeting Papers; Reports - Research
Education Level: N/A
Authoring Institution: N/A
Identifiers: Birnbaum 3 Parameter Logistic Model; Distributions (Statistics); Rasch Model
Note: Paper presented at the Annual Meeting of the American Educational Research Association (64th, Boston, MA, April 7-11, 1980).