NotesFAQContact Us
Search Tips
ERIC Number: ED111868
Record Type: Non-Journal
Publication Date: N/A
Pages: 28
Abstractor: N/A
Reference Count: N/A
Some Methodological Considerations in the Testing of Rasch Model Claims.
Ramaswamy, T.
The Rasch item analysis model is supposed to yield norm-free estimates of ability and easiness values, but there are several possible interpretations of the nature and extent of such norm-freeness. One such interpretation was that to involve the scores of one single experimental group of testees which were embedded in four differently skewed distributions of other scores, the testees had been administered easy and hard sets of items for the verifications of person-free and item-free ability estimations. Similarly for the verifications of person-and item-free easiness estimation by the Rasch model, the study involved the formation of four differently skewed sets of items in terms of their proportion right easiness values among which the same single set of experimental items had been embedded. These four sets of items were administered to bright and dumb groups of examinees. The no-guessing and constant discrimination power assumptions of the Rasch model were respectively made to be satisfied by using the cloze test blanks as items and by removing those blanks outside a narrow range of discrimination indices. Because of the possibility that the estimation errors may critically depend on the number of ties at each raw score level making the Rasch estimates of ability and easiness statistically different from one group of examinees or set of items at each of these score levels, a linear prediction model was used with the raw pupil scores or item easiness as predictors. (Author)
Publication Type: Reports - Research
Education Level: N/A
Audience: N/A
Language: N/A
Sponsor: N/A
Authoring Institution: N/A