ERIC Number: ED370351
Record Type: RIE
Publication Date: 1994-Mar
Theory Building: Sample Size and Data-Model Fit.
de Jong, John H. A. L.; Stoyanova, Fellyanka
A study of item response theory in language testing research investigated the influence of sample size on (1) the statistical test of data-model fit and (2) the invariance of parameter estimates. Data were drawn from a 1993 administration of the examination of Dutch as a second language to about a thousand candidates, using results from only the listening comprehension segment. One group of examinees was divided into several randomly assembled subsamples, differing in size. Independent analyses of the subsamples were run to assess sample size influence on output variables. Second, a subset of test items was selected for which statistical model fit could be shown, and stability of data-model fit and invariance of item and person parameters over several randomly drawn subsamples was evaluated. Results indicate that estimates of data-model fit and item and person parameters are highly dependent on sample size, but that estimates will be stable for randomly drawn subsamples from a large sample for which statistical fit can be shown. Implications for theory building in language testing research are discussed. Contains 14 references. (MSE)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Note: Paper presented at the Annual Language Testing Research Colloquium (March 1994).