ERIC Number: ED390945
Record Type: RIE
Publication Date: 1995-Apr
Reference Count: N/A
The Preferability of Constrained Optimization in Determining the Number of Prompts, Modes of Discourse, and Raters in a Direct Writing Assessment.
Parkes, Jay; Suen, Hoi K.
This study demonstrates the advantages of using a constrained optimization algorithm to explore the optimal number of prompts, modes of discourse, and raters for achieving an acceptable level of reliability during a direct writing assessment. Writing samples elicited from 50 college students were rated by 3 graduate students and the scores submitted to a generalizability analysis (G-study). The variance components estimated in the G-study were then used in a branch-and-bound integer programming procedure to determine the optimal number of raters, modes, and prompts to produce a reliable writing assessment. Four different scenarios were examined to show how the optimal answer changes based on the priorities of the measurement situation. (Contains 2 tables, 3 figures, and 24 references.) (Author/SLD)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Identifiers: Branch and Bound Method
Note: Paper presented at the Annual Meeting of the National Council on Measurement in Education (San Francisco, CA, April 19-21, 1995).