ERIC Number: ED327546
Record Type: RIE
Publication Date: 1989
Reference Count: N/A
Stability of IRT b-Values over Time and Position.
Sykes, Robert C.
An analysis-of-covariance methodology was used to investigate whether there were population differences between tryout and operational Rasch item b-values relative to differences between pairs of item response theory (IRT) b-values from consecutive operational item administrations. This methodology allowed the evaluation of whether any such differences could be attributed to differences in item position after controlling for the background effect of scale drift on b-values. Usable real or scored items in a large licensure examination were selected for study. The final sample consisted of 406 administration difference pairs. For each item, a record was created containing: (1) administration dates; (2) item positions for several administrations ending with July 1987; (3) classification code; (4) type of administration (real or tryout); and (5) b-value and standard error. The block-wise placement of tryout items had no apparent effect on their b-values. Neither differences in booklet nor test item locations were significantly associated with differences in b-values across consecutive pairs of administrations. A small but significant increase in b-value differences in pairs having earlier administration (between February 1981 and July 1986) suggested that scale drift increased over this period. Three data tables and eight figures are included. (Author/SLD)
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: N/A
Identifiers: B Values; Invariance Principle; Item Position (Tests); Rasch Model